CHAPTER CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith, wecan CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3 CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp CHAPTER INTELLIGENT AGENTS Inwhichwediscussthenatureofagents,perfectorotherwise,thediversityofenvironments, andtheresultingmenagerie ofagenttypes. Chapter 1 identified the concept of rational agents as central to our approach to artificial intelligence. Inthischapter,wemakethisnotionmoreconcrete. Wewillseethattheconcept ofrationality canbeappliedtoawidevarietyofagentsoperating inanyimaginable environment. Ourplan inthisbook istousethis concept todevelop asmallset ofdesign principles forbuilding successful agents—systems thatcanreasonably becalledintelligent. Webegin by examining agents, environments, and the coupling between them. The observation that some agents behave better than others leads naturally to the idea of a rational agent—one that behaves as well as possible. How well an agent can behave depends on the natureoftheenvironment;someenvironmentsaremoredifficultthanothers. Wegiveacrude categorization ofenvironments andshowhowproperties ofanenvironment influencethedesignofsuitable agentsforthatenvironment. Wedescribeanumberofbasic“skeleton” agent designs, whichwefleshoutintherestofthebook. 2.1 Agents and Environments Environment Anagentisanything thatcanbeviewedasperceiving itsenvironmentthrough sensorsand Sensor actinguponthatenvironmentthroughactuators. Thissimpleideaisillustratedin Figure2.1. Actuator A human agent has eyes, ears, and other organs for sensors and hands, legs, vocal tract, and so on for actuators. A robotic agent might have cameras and infrared range finders for sensors and various motors for actuators. A software agent receives file contents, network packets, andhumaninput(keyboard/mouse/touchscreen/voice) assensory inputsandactson the environment by writing files, sending network packets, and displaying information or generating sounds. Theenvironment could beeverything—the entire universe! Inpractice it isjustthatpartoftheuniversewhosestatewecareaboutwhendesigningthisagent—thepart thataffectswhattheagentperceivesandthatisaffectedbytheagent’s actions. Percept We use the term percept to refer to the content an agent’s sensors are perceiving. An Percept sequenc◮e agent’sperceptsequenceisthecompletehistoryofeverything theagenthaseverperceived. Ingeneral, anagent’s choice ofaction atanygiven instant candepend onitsbuilt-in knowledge and on the entire percept sequence observed to date, but not on anything it hasn’t perceived. By specifying the agent’s choice of action for every possible percept sequence, we have said more or less everything there is to say about the agent. Mathematically speak Agentfunction ing, wesay that an agent’s behavior is described by the agent function that maps any given perceptsequence toanaction. Section2.1 Agentsand Environments 37 Agent Sensors Actuators Environment Percepts ? Actions Figure2.1 Agentsinteractwithenvironmentsthroughsensorsandactuators. We can imagine tabulating the agent function that describes any given agent; for most agents, this would be a very large table—infinite, in fact, unless we place a bound on the lengthofperceptsequenceswewanttoconsider. Givenanagenttoexperimentwith,wecan, in principle, construct this table by trying out all possible percept sequences and recording whichactionstheagentdoesinresponse.1 Thetableis,ofcourse,anexternalcharacterization of the agent. Internally, the agent function for an artificial agent will be implemented by an agent program. It is important to keep these two ideas distinct. The agent function is an Agentprogram abstract mathematical description; the agent program is a concrete implementation, running withinsomephysicalsystem. To illustrate these ideas, we use a simple example—the vacuum-cleaner world, which consists of a robotic vacuum-cleaning agent in a world consisting of squares that can be either dirty or clean. Figure 2.2 shows a configuration with just two squares, A and B. The vacuum agent perceives which square it is in and whether there is dirt in the square. The agentstartsinsquare A. Theavailable actionsaretomovetotheright,movetotheleft,suck up the dirt, or do nothing.2 One very simple agent function is the following: if the current square is dirty, then suck; otherwise, move to the other square. A partial tabulation of this agent function is shown in Figure 2.3 and an agent program that implements it appears in Figure2.8onpage49. Looking at Figure 2.3, we see that various vacuum-world agents can be defined simply ◭ byfillingintheright-handcolumninvariousways. Theobviousquestion,then,isthis: What is the right way to fill out the table? In other words, what makes an agent good or bad, intelligent orstupid? Weanswerthesequestions inthenextsection. 1 Iftheagentusessomerandomizationtochoose itsactions, thenwewouldhavetotryeachsequence many timestoidentifytheprobabilityofeachaction. Onemightimaginethatactingrandomlyisrathersilly,butwe showlaterinthischapterthatitcanbeveryintelligent. 2 Inarealrobot,itwouldbeunlikelytohaveanactionslike“moveright”and“moveleft.” Insteadtheactions wouldbe“sp