What Is An Agent ? Chapter2 Intelligent Agents 2 20070308 chap2 1 20070308 chap2 What Is An Agent ? (cont.-1) What Is An Agent ? (cont.-2) An agent interacts with its environments through • Agent function (abstract mathematical description) sensors and actuators. that maps any given percept sequence to an action. • Perceiving through sensors - human agent: eyes, ears, etc. - robot agent: cameras, infrared, etc. • Agent program (concrete implementation) - software agent: receiving keystrokes, file contents, that implements the agent function, network packets, etc. running on the physical architecture to produce f . • Acting through actuators - human agent: hands, legs, mouth, etc. • A rational agent is one that does the right thing. - robot agent: arms, motors, etc. - software agent: displaying on the screen, sending network packets, etc. 3 4 20070308 chap2 20070308 chap2 Vacuum-cleaner world Vacuum-cleaner world (cont.) • agent function • agent program Percepts: location and contents, e.g., [ A, Dirty ] Actions: Left, Right, Suck, NoOp What is the right function? Can it be implemented in a small agent program? 5 6 20070308 chap2 20070308 chap2 1
Performance Measure Rational Agents The right action is the one that will cause What is rational depends on the agent to be most successful . • Performance measure - degree of success • Percept sequence to date • The measure should be objective. • The agent's knowledge about the environment • How does one evaluate success? • Actions that can be performed by the agent • When does one evaluate success? • Example: to vacuum a dirty floor Definition of a rational agent the cleanness of the floor For each possible percept sequence, rational agent the amount of dirt cleaned up should select an action that is expected to maximize its the amount of electricity consumed performance measure, given the evidence provided by the amount of noise generated the percept sequence and whatever built-in knowledge total time and effort spent the agent has. performance over a short/long time 7 8 20070308 chap2 20070308 chap2 Rational Agents (cont.) Specifying the Task Environment • rationality ≠ omniscience ( 全知 ) e.g. crossing the road and be flattened PEAS (Performance, Environment, Actuators, Sensors) • rationality ≠ perfection Rationality maximizes expected performance, e.g. Automated Taxi while perfection maximizes actual performance . • rationality = exploration + learning + autonomy e.g. lowly dung beetles, female sphex wasp Information gathering/exploration -- To maximize future rewards Learn from percepts -- To extend prior knowledge Agent autonomy -- To compensate for incorrect/partial prior knowledge 9 10 20070308 chap2 20070308 chap2 Specifying the Task Environment (cont.) Properties of Task Environments e.g. Medical diagnosis system Performance measure?? Healthy patients, minimize costs Solitaire Backgammom Internet Taxi ( 西洋雙陸棋戲 ) shopping ( 接龍 ) Environment?? Patient, hospital, staff Observable?? Deterministic?? Actuators?? Display questions, tests, treatments, diagnoses, referrals Episodic?? Sensors?? Keyboard entry of symptoms, findings, patient’s answer Static?? Discrete?? Single-agent?? 11 12 20070308 chap2 20070308 chap2 2
Properties of Task Environments ( cont.-1 ) Properties of Task Environments ( cont.-2 ) Fully vs. partially observable : an environment is full observable when the Fully vs. partially observable : an environment is full observable when the sensors can detect all aspects that are relevant to the choice of action. sensors can detect all aspects that are relevant to the choice of action. Solitaire Backgammom Internet Taxi Solitaire Backgammom Internet Taxi shopping shopping ( 西洋雙陸棋戲 ) ( 西洋雙陸棋戲 ) ( 接龍 ) ( 接龍 ) Observable?? Observable?? FULL FULL PARTIAL PARTIAL Deterministic?? Deterministic?? Episodic?? Episodic?? Static?? Static?? Discrete?? Discrete?? Single-agent?? Single-agent?? 13 14 20070308 chap2 20070308 chap2 Properties of Task Environments ( cont.-3 ) Properties of Task Environments ( cont.-4 ) Deterministic vs. stochastic : if the next environment state is completely Deterministic vs. stochastic : if the next environment state is completely determined by the current state the executed action then the environment is determined by the current state the executed action then the environment is deterministic. If it is deterministic except for actions of other agents, we say the deterministic. If it is deterministic except for actions of other agents, we say the environment is strategic. environment is strategic. Solitaire Backgammom Internet Taxi Solitaire Backgammom Internet Taxi shopping shopping ( 接龍 ) ( 西洋雙陸棋戲 ) ( 接龍 ) ( 西洋雙陸棋戲 ) Observable?? FULL FULL PARTIAL PARTIAL Observable?? FULL FULL PARTIAL PARTIAL Deterministic?? Deterministic?? YES NO YES NO Episodic?? Episodic?? Static?? Static?? Discrete?? Discrete?? Single-agent?? Single-agent?? 15 16 20070308 chap2 20070308 chap2 Properties of Task Environments ( cont.-5 ) Properties of Task Environments ( cont.-6 ) Episodic vs. sequential : In an episodic environment the agent ’ s experience Episodic vs. sequential : In an episodic environment the agent ’ s experience can be divided into atomic steps where the agents perceives and then performs can be divided into atomic steps where the agents perceives and then performs a single action. The choice of action depends only on the episode itself . a single action. The choice of action depends only on the episode itself . Solitaire Backgammom Internet Taxi Solitaire Backgammom Internet Taxi ( 西洋雙陸棋戲 ) shopping ( 西洋雙陸棋戲 ) shopping ( 接龍 ) ( 接龍 ) Observable?? FULL FULL PARTIAL PARTIAL Observable?? FULL FULL PARTIAL PARTIAL Deterministic?? YES NO YES NO Deterministic?? YES NO YES NO Episodic?? Episodic?? NO NO NO NO Static?? Static?? Discrete?? Discrete?? Single-agent?? Single-agent?? 17 18 20070308 chap2 20070308 chap2 3
Properties of Task Environments ( cont.-7 ) Properties of Task Environments ( cont.-8 ) Static vs. dynamic : If the environment can change while the agent is choosing Static vs. dynamic : If the environment can change while the agent is choosing an action, the environment is dynamic. Semi-dynamic if the agent ’ s performance an action, the environment is dynamic. Semi-dynamic if the agent ’ s performance changes even when the environment remains the same. changes even when the environment remains the same. Solitaire Backgammom Internet Taxi Solitaire Backgammom Internet Taxi shopping shopping ( 西洋雙陸棋戲 ) ( 西洋雙陸棋戲 ) ( 接龍 ) ( 接龍 ) Observable?? FULL FULL PARTIAL PARTIAL Observable?? FULL FULL PARTIAL PARTIAL Deterministic?? YES NO YES NO Deterministic?? YES NO YES NO Episodic?? NO NO NO NO Episodic?? NO NO NO NO Static?? Static?? YES YES SEMI NO Discrete?? Discrete?? Single-agent?? Single-agent?? 19 20 20070308 chap2 20070308 chap2 Properties of Task Environments ( cont.-9 ) Properties of Task Environments ( cont.-10 ) Discrete vs. continuous : This distinction can be applied to the state of the Discrete vs. continuous : This distinction can be applied to the state of the environment, the way time is handled and to the percepts/actions of the agent. environment, the way time is handled and to the percepts/actions of the agent. Solitaire Backgammom Internet Taxi Solitaire Backgammom Internet Taxi shopping shopping ( 接龍 ) ( 西洋雙陸棋戲 ) ( 接龍 ) ( 西洋雙陸棋戲 ) Observable?? FULL FULL PARTIAL PARTIAL Observable?? FULL FULL PARTIAL PARTIAL Deterministic?? YES NO YES NO Deterministic?? YES NO YES NO Episodic?? NO NO NO NO Episodic?? NO NO NO NO Static?? YES YES SEMI NO Static?? YES YES SEMI NO Discrete?? Discrete?? YES YES YES NO Single-agent?? Single-agent?? 21 22 20070308 chap2 20070308 chap2 Properties of Task Environments ( cont.-11 ) Properties of Task Environments ( cont.-12 ) Single vs. multi-agent: Does the environment contain other agents who Single vs. multi-agent: Does the environment contain other agents who are also maximizing some performance measure that depends on the are also maximizing some performance measure that depends on the current agent ’ s actions? current agent ’ s actions? Solitaire Backgammom Internet Taxi Solitaire Backgammom Internet Taxi ( 西洋雙陸棋戲 ) shopping ( 西洋雙陸棋戲 ) shopping ( 接龍 ) ( 接龍 ) Observable?? FULL FULL PARTIAL PARTIAL Observable?? FULL FULL PARTIAL PARTIAL Deterministic?? YES NO YES NO Deterministic?? YES NO YES NO Episodic?? NO NO NO NO Episodic?? NO NO NO NO Static?? YES YES SEMI NO Static?? YES YES SEMI NO Discrete?? YES YES YES NO Discrete?? YES YES YES NO Single-agent?? YES NO NO NO Single-agent?? 23 24 20070308 chap2 20070308 chap2 4
Recommend
More recommend
