Intelligent AgentsIntelligent Agente.g., HumansNotion of an Artificial AgentSlide 5Agents and environmentsSlide 7Vacuum Cleaner WorldVacuum Agent FunctionThe vacuum-cleaner worldThe concept of rationalityRationalitySlide 13Slide 14EnvironmentsSlide 16Environment typesSlide 18Slide 19Slide 20Slide 21Slide 22Slide 23Slide 24Slide 25Slide 26Slide 27Slide 28Slide 29Slide 30Agent typesAgent typesSlide 33Simple reflex agentsSlide 35Simple reflex agentModel-based reflex agentSlide 38Goal-based agentsUtility-based agentsLearning agentsLearning AgentsSummary: Intelligent AgentsIntelligent AgentsIntelligent AgentsRussell and Norvig: Chapter 2CMSC421 – Fall 2006Intelligent AgentDefinition: An intelligent agent perceives its environment via sensors and acts rationally upon that environment with its actuators. environmentagent?sensorsactuatorsperceptsactionsSensors: Eyes (vision), ears (hearing), skin (touch), tongue (gustation), nose (olfaction), neuromuscular system (proprioception)Percepts: At the lowest level – electrical signals After preprocessing – objects in the visual field (location, textures, colors, …), auditory streams (pitch, loudness, direction), …Actuators: limbs, digits, eyes, tongue, …Actions: lift a finger, turn left, walk, run, carry an object, …e.g., HumansNotion of an Artificial Notion of an Artificial AgentAgentenvironmentagent?sensorsactuatorslaser range findersonarstouch sensorsNotion of an Artificial Notion of an Artificial AgentAgentenvironmentagent?sensorsactuatorsAgents and environmentsAgents include human, robots, softbots, thermostats, etc.The agent function maps percept sequence to actionsAn agent can perceive its own actions, but not always it effects.f : P*  AAgents and environmentsThe agent function will internally be represented by the agent program.The agent program runs on the physical architecture to produce f.Vacuum Cleaner WorldEnvironment: square A and BPercepts: location and content, e.g. [A, Dirty]Actions: Left, Right, Suck, NoOpVacuum Agent FunctionPercept Sequence Action[A, Clean] Right[A, Dirty] Suck[B, Clean] Left[B, Dirty] Suck[A, Clean], [A, Clean]Right[A, Clean], [A, Dirty] Suck…The vacuum-cleaner worldfunction REFLEX-VACUUM-AGENT ([location, status]) return an actionif status == Dirty then return Suckelse if location == A then return Rightelse if location == B then return LeftWhat is the right function? Can it be implemented in a small agent program?The concept of rationalityA rational agent is one that does the right thing.Every entry in the table is filled out correctly.What is the right thing?Approximation: the most successful agent.Measure of success?Performance measure should be objectiveE.g. the amount of dirt cleaned within a certain time.E.g. how clean the floor is.…Performance measure according to what is wanted in the environment instead of how the agents should behave.RationalityWhat is rational at a given time depends on four things:Performance measure,Prior environment knowledge,Actions,Percept sequence to date (sensors). Definition: A rational agent chooses whichever action maximizes the expected value of the performance measure given the percept sequence to date and prior environment knowledge.RationalityRationality  omniscienceAn omniscient agent knows the actual outcome of its actions.Rationality  perfectionRationality maximizes expected performance, while perfection maximizes actual performance.Rationality The proposed definition requires:Information gathering/explorationTo maximize future rewardsLearn from perceptsExtending prior knowledgeAgent autonomyCompensate for incorrect prior knowledgeEnvironmentsTo design a rational agent we must specify its task environment.PEAS description of the environment:Performance EnvironmentActuatorsSensorsEnvironmentsE.g. Fully automated taxi:PEAS description of the environment:Performance Safety, destination, profits, legality, comfortEnvironmentStreets/freeways, other traffic, pedestrians, weather,, …ActuatorsSteering, accelerating, brake, horn, speaker/display,…SensorsVideo, sonar, speedometer, engine sensors, keyboard, GPS, …Environment typesSolitaire Backgammom Intenet shopping TaxiObservable??Deterministic??Episodic??Static??Discrete??Single-agent??Environment typesSolitaire Backgammon Internet shopping TaxiObservable??Deterministic??Episodic??Static??Discrete??Single-agent??Fully vs. partially observable: an environment is full observable when the sensors can detect all aspects that are relevant to the choice of action.Environment typesSolitaire Backgammon Internet shopping TaxiObservable?? PARTIAL FULL PARTIAL PARTIALDeterministic??Episodic??Static??Discrete??Single-agent??Fully vs. partially observable: an environment is full observable when the sensors can detect all aspects that are relevant to the choice of action.Environment typesSolitaire Backgammon Internet shopping TaxiObservable?? PARTIAL FULL PARTIAL PARTIALDeterministic??Episodic??Static??Discrete??Single-agent??Deterministic vs. stochastic: if the next environment state is completelydetermined by the current state the executed action then the environment isdeterministic.Environment typesSolitaire Backgammon Internet shopping TaxiObservable?? PARTIAL FULL PARTIAL PARTIALDeterministic?? YES NO YES NOEpisodic??Static??Discrete??Single-agent??Deterministic vs. stochastic: if the next environment state is completelydetermined by the current state the executed action then the environment isdeterministic.Environment typesSolitaire Backgammon Internet shopping TaxiObservable?? PARTIAL FULL PARTIAL PARTIALDeterministic?? YES NO YES NOEpisodic??Static??Discrete??Single-agent??Episodic vs. sequential: In an episodic environment the agent’s experiencecan be divided into atomic steps where the agents perceives and then performsA single action. The choice of action depends only on the episode itselfEnvironment typesSolitaire Backgammon Internet shopping TaxiObservable?? PARTIAL FULL PARTIAL PARTIALDeterministic?? YES NO YES NOEpisodic?? NO NO NO NOStatic??Discrete??Single-agent??Episodic vs. sequential: In an episodic environment the agent’s experiencecan be divided into atomic steps where the agents perceives and then performsA single action. The choice of action depends only on the episode itselfEnvironment typesSolitaire Backgammon Internet shopping TaxiObservable?? PARTIAL