1Administrative IssuesLogin into learn usc edu and make sureLogin into learn.usc.edu and make sure that CSCI561a is listed as one of your courses. Web page:Web page: http://www-scf.usc.edu/~csci561b/ http://den.usc.eduActing Humanly: The Full Turing Test•Problem:1) Turing test is not reproducible, constructive, and amenable to mathematic analysis. 2) What about physical interaction with interrogator and environment?Trap door2Last time: The Turing Testhttp://www.ai.mit.edu/projects/infolab/http://aimovie.warnerbros.comThis time: OutlineIntelligent Agents (IA)Intelligent Agents (IA) Environment types IA Behavior IA StructureIA TIA Types3Attributes of Intelligent Behavior Think and reason Use reason to solve problems Learn or understand from experience Acquire and apply knowledge Exhibit creativity and imagination Deal with complex or perplexing situations Respond quickly and successfully to new it tisituations. Recognize the relative importance of elements in a situation Handle ambiguous, incomplete,or erroneous informationIntelligent AgentsInterfaceTutorsSearchAgentsPresentationAgentsNetworkNavigationAtUserInterfaceAgentsInformationManagementAgentsInformationBrokersAgentsRole-PlayingAgentsInformationFilters4What is an (Intelligent) Agent?An overused overloaded andAn over-used, over-loaded, and misused term. Anything that can be viewed asperceiving its environmentpg acting upon that environment What is an (Intelligent) Agent?PAGE (Percepts Actions GoalsPAGE (Percepts, Actions, Goals, Environment) Task-specific & specialized: well-defined goals and environmentg5Intelligent Agents and Artificial Intelligence Example: Human mind as network of pthousands or millions of agents working in parallel.AgencysensorseffectorsAgent TypesAgent research fall into two main strands:Agent research fall into two main strands: Distributed Artificial Intelligence (DAI) –Multi-Agent Systems (MAS) (1980 – 1990) Much broader notion of "agent" (1990’s – present)6Rational AgentsHo to design this?EnvironmentApercepts?SensorsHow to design this?AgentactionsEffectorsA Windshield Wiper AgentHowdowedesignanagentthatcanwipeHow do we design an agent that can wipe the windshields when needed? Goals? Percepts ?S?Sensors? Effectors ? Actions ? Environment ?7Grand ChallengeAutonomous DrivingAutonomous DrivingInteracting AgentsCollision Avoidance Agent (CAA) Goals: Avoid running into obstacles Percepts ? Sensors? Effectors ?Actions ? Environment: Freeway8Interacting AgentsLane Keeping Agent (LKA)• Goals: Stay in current lane•Percepts ?•Sensors?•Effectors ?Effectors ?•Actions ?• Environment: FreewayConflict Resolution by Action Selection Agents•Override:Override:• Arbitrate:• Compromise:• Challenges:9The Right Thing = The Rational ActionRational Action:The action that maximizes theRational Action:The action that maximizes the expected value of the performance measure given the percept sequence to date Rational = Best ? Rational = Optimal ?Rti l O i i ?(H i ttlRational = Omniscience ? (Having total knowledge) Rational = Clairvoyant ? (The sixth sense) Rational = Successful ?Behavior and performance of IAsPerception(sequence) toActionPerception(sequence) to ActionMapping:f : P* → A Ideal mapping:10Look up tableobstaclesensorDistanceActtionagentsensorDistanceActtion10 No action5Turn left 305Turn left 30 degrees2StopClosed formOutput (degree of rotation) =Output (degree of rotation) = F(distance)11Behavior and performance of IAsPerformance measure:Performance measure:(d f)At(degree of)Autonomy:How is an Agent different from other software?Agents areautonomousAgents are autonomous, Agents contain some level of intelligence, Agents don't only act reactively, but sometimes also proactively12How is an Agent different from other software? Agents have social ability, Agents may cooperate with other agents Agents may migrate from one system to anotherEnvironment Types Characteristics Accessible vs. inaccessible Deterministic vs. nondeterministic Episodic vs. nonepisodic (Sequential)13Environment Types Characteristics Hostile vs. friendly Static vs. dynamic Discrete vs. continuous Environment typesEnvironmentAccessiDeterminisEpisodicStaticDiscreteEnvironmentAccessibleDeterministicEpisodicStaticDiscreteOperating SystemVirtual RealityOffiOffice EnvironmentMars14Structure of Intelligent AgentsAgent = architecture + programAgent = architecture + program Agent program: the implementation of f : P* → A, the agent’s perception-action mappingfunction: Skeleton-Agent(Percept) returns Actionmemory ← UpdateMemory(memory, Percept)Action ← ChooseBestAction(memory)memory ← UpdateMemory(memory, Action)return ActionUsing a look-up-table to encode f : P* → A Example: Collision Avoidancep Sensors: 3 proximity sensors Effectors: Steering Wheel, Brakes How to generate? How large? How to select action?agentobstaclesensors15Using a look-up-table to encode f : P* → Abt l Example: Collision Avoidance Sensors: 3 proximity sensors Effectors: Steering Wheel, BrakesagentobstaclesensorsUsing a look-up-table to encode f : P* → AHow large:How large: How to select action? Is it an autonomous agent? (by using the look up table)16Agent types Reflex agents g Reactive: No memory Reflex agents with internal states Goal-based agents Goal information needed to make decisionAgent types Utility-based agentsyg Learning Agent17Reflex agentsQuestionDesign a group of mobile robots thatDesign a group of mobile robots that stay together and move around using reactive (reflex) agents?18Reflex agents w/ state (model-based reflex agent)Goal-based agents19Utility-based agentsLearning agentsPerformance standardPerformance elementLearning elementCriticChangesldfeedbackLearningProblem generatorKnowledgeLearning goal20Information agents Manage the explosive growth of information. Information agentsExamples:Examples: BargainFinder comparison shops among Internet stores for CDs FIDO the Shopping Doggie (out of service) Internet Softbot infers which internet facilities (finger, ftp, gopher) to use and when from high-level search requests. Challenge: ontologies for annotating Web pages (eg, SHOE).21Example: ALADDIN projectAutonomous Learning
View Full Document