AI is pretty hard stuff{Agents and Environments}Agents as Mappings{Vacuum-cleaner world}{A vacuum-cleaner agent}More Examples of Artificial Agents{Rationality}Rational Agents{PEAS}{PEAS for an Automated Taxi}{Internet shopping agent}Environment Types{Environment Types}{Environment types}{Environment types}{Environment types}{Environment types}{Environment types}{Environment types}{Environment types: Vacuum-Cleaner}{Environment types}Agent ProgramsDifferent Types of AgentsA Reflex Taxi-Driver AgentSimple Reflex AgentsReflex Taxi-Driver Agent with StateReflex Agents with Internal StateReflex Taxi-Driver Agent with StateA Goal-based Taxi-Driver AgentGoal-based AgentsUtility-based Taxi-Driver Agent{Utility-based Agents}{Learning Agents}Intelligent AgentsReadings: Chapter 2 ofRussell & Norvig.Artificial Intelligence – p.1/34AI is pretty hard stuffI went to the grocery store, I saw the milk on the shelf and Ibought it.What did I buy?The milk?The shelf?The store?An awful lot of knowledge of the world is needed to answersimple questions like this one.Artificial Intelligence – p.2/34Agents and EnvironmentsAn agent is a system that perceives its environment throughsensors and acts upon that environment through effectors.AgentSensorsActuatorsEnvironmentPerceptsActions?Agents include humans, robots, softbots, thermostats, etc.Artificial Intelligence – p.3/34Agents as MappingsAn agent can be seen as a mapping between perceptsequences and actions.Agent : Percept∗−→ Action∗The less an agents relies on its built-in knowledge, asopposed to the current percept sequence, the moreautonomous it is.A rational agent is an agent whose acts try to maximizesomeperformance measure.Artificial Intelligence – p.4/34Vacuum-cleaner worldA BPercepts: location and contents, e.g., [A, Dirty]Actions: Left, Right, Suck, N oOpArtificial Intelligence – p.5/34A vacuum-cleaner agentPercept sequence Action[A, Clean] Right[A, Dirty] Suck[B, Clean] Left[B, Dirty] Suck[A, Clean], [A, Clean] Right[A, Clean], [A, Dirty] Suck......function REFLEX-VACUUM-AGENT([location,status]) returnsactionif status = Dirty then return Suckelse if location = A then return Rightelse if location = B then return LeftArtificial Intelligence – p.6/34More Examples of Artificial AgentsAgent Type Percepts Actions Goals EnvironmentMedical diagnosissystemSymptoms,findings, patient’sanswersQuestions, tests,treatmentsHealthy patient,minimize costsPatient, hospitalSatellite imageanalysis systemPixels of varyingintensity, colorPrint acategorization ofsceneCorrectcategorizationImages fromorbiting satellitePart-picking robot Pixels of varyingintensityPick up parts andsort into binsPlace parts incorrect binsConveyorbeltwith partsRefinery controller Temperature,pressure readingsOpen, closevalves; adjusttemperatureMaximize purity,yield, safetyRefineryInteractive EnglishtutorTyped words Print exercises,suggestions,correctionsMaximizestudent’s score ontestSet of studentsArtificial Intelligence – p.7/34RationalityWhat is the right function?Can it be implemented in a small agent program?Fixed performance measure evaluates the environmentsequenceone point per square cleaned up in time T ?one point per clean square per time step, minus oneper move?penalize for > k dirty squares?Rational 6= omniscientRational 6= successfulRational =⇒ exploration, learning, autonomyArtificial Intelligence – p.8/34Rational AgentsThe rationality of an agent depends onthe performance measure defining the agent’s degree ofsuccessthe percept sequence, the sequence of all the thingsperceived by the agentthe agent’s knowledge of the environmentthe actions that the agent can performFor each possible percept sequence, anideal rational agentdoes whatever possible to maximize its performance, basedon the percept sequence and its built-in knowledge.Artificial Intelligence – p.9/34PEASTo design a rational agent, we must specify the taskenvironmentConsider, e.g., the task of designing an automated taxi:Performance measure??Environment??Actuators??Sensors??Artificial Intelligence – p.10/34PEAS for an Automated TaxiThe task of designing an automated taxi:Performance measure?? safety, destination, profits,legality, comfort, . . .Environment?? US streets/freeways, traffic,pedestrians, weather, . . .Actuators?? steering, accelerator, brake, horn,speaker/display, . . .Sensors?? video, accelerometers, gauges, enginesensors, keyboard, GPS, . . .Artificial Intelligence – p.11/34Internet shopping agentPerformance measure??Environment??Actuators??Sensors??Artificial Intelligence – p.12/34Environment TypesWith respect to an agent, an environment may, or may not,be:accessible: the agent’s sensors detect all aspectsrelevant to the choice of action;deterministic: the next state is completely determined bythe current state and the actions selected by the agent;episodic: the agent’s experience is divided into“episodes”; the quality of the agent’s actions does notdepend on previous episodes;static: it does not change while the agent is deliberating;discrete: there are a limited number of distinct, clearlydefined percepts and actions.Artificial Intelligence – p.13/34Environment TypesSolitaire Backgammon E-shopping TaxiObservable??Deterministic??Episodic??Static??Discrete??Single-agent??Artificial Intelligence – p.14/34Environment typesSolitaire Backgammon E-shopping TaxiObservable?? Yes Yes No NoDeterministic??Episodic??Static??Discrete??Single-agent??Artificial Intelligence – p.15/34Environment typesSolitaire Backgammon E-shopping TaxiObservable?? Yes Yes No NoDeterministic?? Yes No Partly NoEpisodic??Static??Discrete??Single-agent??Artificial Intelligence – p.16/34Environment typesSolitaire Backgammon E-shopping TaxiObservable?? Yes Yes No NoDeterministic?? Yes No Partly NoEpisodic?? No No No NoStatic??Discrete??Single-agent??Artificial Intelligence – p.17/34Environment typesSolitaire Backgammon E-shopping TaxiObservable?? Yes Yes No NoDeterministic?? Yes No Partly NoEpisodic?? No No No NoStatic?? Yes Semi Semi NoDiscrete??Single-agent??Artificial Intelligence – p.18/34Environment typesSolitaire Backgammon E-shopping TaxiObservable?? Yes Yes No NoDeterministic?? Yes No Partly NoEpisodic?? No No No NoStatic?? Yes Semi Semi NoDiscrete?? Yes Yes Yes NoSingle-agent??Artificial Intelligence – p.19/34Environment typesSolitaire Backgammon E-shopping TaxiObservable?? Yes Yes No NoDeterministic?? Yes No Partly NoEpisodic?? No No No NoStatic?? Yes Semi Semi