17/10/2012 Outline Special Topics in AI: Intelligent Agents and Multi-Agent Systems • Course Presentation – Aims, schedule, exam modalities Course Presentation and Introduction • Intelligent agents – AI, Intelligent agents, Rationality • Multi-Agent Systems – Main features, techniques, applications Alessandro Farinelli Lecture Material Course Organization Artificial Intelligence – A Modern Approach Wed 17th Oct. 15:30 -- 17:30 Room M; by Stuart Russell - Peter Norvig Tue 23rd Oct. 15:30 -- 17:30; Room H Tue 30th Oct. 15:30 -- 17:30; Room H Mon. 5th Nov. 16:00 -- 18:00; Sala Verde An Introduction to Multiagent Systems Tue. 13th Nov. 15:30 -- 17:30; Room H by Michael Wooldridge Tue. 20th Nov. 15:30 -- 17:30; Room H Tue. 27th Nov. 15:30 -- 17:30; Room H Multiagent Systems. 2nd Edition. Tue. 4th Dec. 15:30 -- 17:30; Room H Gherard Weiss (Ed.) Tue. 11th Dec. 15:30 -- 17:30; Room H Tue. 18th Dec. 15:30 -- 17:30; Room H Lecture slides and Info: 1
17/10/2012 Course Aim Course Program At the end of this course will be able to: 1. Decentralized Coordination – Modeling Decentralized Coordination as DCOPs 1. Understand main issues and research challenges for Multi-Agent Systems – DCOPs solution techniques (exact and approx.) 2. Market Based Allocation – Decentralized Coordination, Market Based Allocation, Reasoning under uncertainty – Auction Mechanisms, Combinatorial auctions, Sequential 2. Model and solve Decentralized Coordination problems auctions 3. Reasoning under uncertainty – DCOPs (exact and approx. methods) – 3. Understand main models and solution techniques for MDPs, POMDPs decision making under uncertainty – Probabilistic approaches for robot navigation – MDP, POMDPs, Dec-MDPs Exam modalities Outline • Students read, present to the class, and discuss a set of selected papers. • Course Presentation • Student together with instructor choose papers – Aims, schedule, exam modalities – Topics: Decentralized optimization, Market-Based Allocation, • Intelligent agents Reasoning under uncertainty (robotics) – AI, Intelligent agents, Rationality • Presentation: • Multi-Agent Systems – From 45mins to 1 hour + questions – During the last three lessons (4 th 11 th 18 th Dec.) – Main features, techniques, applications 2
17/10/2012 What is AI? Acting Humanly:Turing Test Turing(1950) Computing Machinery and Intelligence • Can machine think? � Can machine behave like humans? • Operational test: the imitation game Problem: not reproducible, constructive or amenable to mathematical analysis Thinking humanly: Cognitive Thinking rationally: Laws of Science thoughts • Cognitive Neuroscience � theories of internal • Normative not descriptive activities of the brains • Problems: – Level of abstraction? Validation ? – Intelligence not always based on logical deliberation • Available theories do not explain human-level – What are the purpose of thinking ? Which thoughts should intelligence I have out of all the ones that I could have 3
17/10/2012 Acting rationally Rational agents • Do the right thing • Agent: entity that perceives and acts – Action that maximizes some measure of performances • Rational agent given current information – A function from percept histories to actions • Thinking should be in service of rational actions – Thinking is not necessary (e.g., blinking reflex) • Correct thinking (inference) does not always result in – For a given class of environments and tasks we seek the rational actions agent with best performance (optimization problem) – Thinking is not sufficient Agents and Environments Rationality • Given a performance measure for environment sequences • Rational agent: chooses actions that maximizes the expected value given percept sequence • RaDonal E omniscient – Perception may not supply all relevant info • Agents: humans, softbots, thermostats, robots, etc. • RaDonal E clairvoyant • Agent function: maps perception histories to actions – Action outcome might be unexpected • Agent program: implements the agent function on • Hence RaDonal E successful the physical architecture • Rational => exploration, learning, autonomy,… • 4
17/10/2012 Agent Types: Simple reflex Agent Agent Types: Goal-Based agents Agent Types: Utility-Based Agent AI (recent) history 5
17/10/2012 AI Exciting Applications Example: Search and Rescue • Game Playing – IBM’s Deep Blue (1997) – Poker (Now) http://webdocs.cs.ualberta.ca/~games/poker/ • Autonomous Control – Google self driving car http://www.ted.com/talks/sebastian_thrun_google_s_driverle ss_car.html • Search and Recue/hostile environments – RoboCup Rescue (http://www.robocuprescue.org/ ) • Human Agent Collectives – Orchid project (http://www.orchid.ac.uk/project-aims/) LabRoCoCo http://labrococo.dis.uniroma1.it/wiki/doku.php Intelligent Agents Outline • Intelligent Agents: rational agent + – Reactivity • Course Presentation – Pro-activeness – Aims, schedule, exam modalities – Social ability � Multi-Agent systems • Intelligent agents – AI, Intelligent agents, Rationality • Multi-Agent Systems Rational Agent Intelligent Agent – Main features, techniques, applications 6
17/10/2012 Multi-Agent Systems MAS Characteristics • (Durfee and Lesser 1989): “loosely coupled network of (K. P. Sycara 1998) problem solvers that interact to solve problems that 1. Each agent has incomplete information or are beyond the individual capabilities or knowledge capabilities for solving the problem and, thus, has of each problem solver “ a limited viewpoint • Problem solvers: Intelligent agents 2. There is no system global control • (John Gage, Sun Microsystems) 3. Data is decentralized “The network is the computer” 4. Computation is asynchronous Example: cooperative foraging Why MAS? • To solve problem that are too large for a single agent – Problem decomposition • To Avoid single point of failure in critical applications – Disaster mitigation/urban search and rescue • To model problem that are naturally described with collectives of autonomous components – Meeting scheduling, Traffic control, Forming coalition of customers, … 7
17/10/2012 Applications of MAS I: Main Research Areas in MAS Games, entertainment and education SDM KR Real Time Strategy (e.g. Starcraft, • MDPs Age of Empires) • ATL � group formation, task assignment, Coordination strategic planning Game Theory • Graphical models • core stability First Person Shooter (e.g. Half Life 2, Splinter Cell) � character interactions MAS Cooperative information gathering Applications of MAS II: Search and Rescue Limited communication Cooperative information gathering Joint work with Stranders, Rogers, Jennings [IJCAI 09] UAVs cooperative image collection 8
17/10/2012 CIG: the model CIG: goal Predictive Uncertainty • Monitor a spatial phenomena Contours • Model: scalar field • Minimise prediction uncertainty – Two spatial dimensions – One temporal dimension • Given a measure here what is my uncertainty over there • Tools: – Gaussian process • Estimate uncertainty – Entropy Measures • Measure information CIG: Performance measure and CIG: Demo interactions U U U 1 3 2 H ( X ) H ( X 2 X | ) H ( X | X , X ) 1 1 1 2 3 H ( X , X , X ) H ( X ) H ( X | X ) H ( X | X , X ) = + + 1 2 3 1 2 1 3 1 2 � U U U U + + = 1 2 3 i 9
17/10/2012 CIC: Task utility Cooperative Image Collection Task completion Task Assignment for UAVs Priority Joint work with: Delle Fave, Rogers, Video Streaming Jennings Urgency Coordination First assigned UAVs reaches task Interest points Last assigned UAVs leaves task (consider battery life) 38 CIC: UAVs Demo CIC: Interactions UAV PDA 2 2 X PDA 2 1 T 1 T U 2 2 U 1 UAV 1 U 3 T 3 X PDA 1 3 10
17/10/2012 Applications of MAS III: Electricity markets Energy management Mechanism design and Energy Intelligent agents for the smart grid trading • Force demand to follow supply Home Energy management • Agents to decide load scheduling and storage Collective energy trading • Buy and sell energy as collectives Electricity markets Electricity markets PeakLoad Baseload Baseload: Carried by baseload stations with low cost generation, efficiency and safety Peakload: Carried by expensive, carbon-intensive peaking plants generators 11
17/10/2012 Electricity group purchasing Electricity Group Purchasing • Virtual Electricity Consumer (VEC): A group of consumers • Allow group purchasing among electricity that act in the market as a single energy consumer. consumers • Very popular successful cases – Groupon, Groupalia – UK Labour party initiative on collective electricity purchase + Social networks Group synergies • Social networks to support the • Traditional group purchasing based on group size VEC formation and • Group synergy: complementary energy management restrictions • Look for potential partners • Flattened demand => Better prices through its contacts • VECs of friends of friends 12
Recommend
More recommend
