MultiAgent Systems MAS Tom Holvoet Dept Computer Science

Multi-Agent Systems (MAS) Tom Holvoet Dept. Computer Science, KULeuven Research group on Distributed Systems and Computer Networks

Topic 1: Introduction background n course material n goals / contents /. . . n class sessions / practical work n examination n practical information n course overview n … n

MAS course: objectives / content n MAS course Software Engineering using Multi-Agent Systems ¨ situate MAS in “mainstream software engineering” n mostly distributed software engineering n just a few topics in this course, and even then, just fragments of these topics ¨ ¨ ¨ ¨ general perspective typical agent architectures interaction communication ant-like systems … “Agents are 99% computer science, and 1% A. I. ” (Etzioni)

MAS course: objectives / content n After this course, you should be able to… Situate multi-agent systems in a software engineering perspective ¨ Judging the usefulness of a multi-agent system for distributed problem solving ¨ Make well-founded decisions in defining multi-agent system software architectures, including ¨ n n n Agents Agent behavior architecture Coordination schemes ¨ Message-based vs. envrironment-centric coordination taking into account non-functional requirements of scalability, flexibility, adaptability

MAS prerequisites n object-oriented programming n n n exercises terminology software development n basic terminology of distributed systems / AI n helps: n software engineering in general ¨ UP, requirements, analysis, OOD, design patterns

MAS: classes, exercises, practical work n officially: n n n 19. 5 hours 10 hours lectures exercise sessions planning in practice: n +- 7 lectures (1. 5 h) ¨ ¨ n “theory” discussion 4 exercise sessions (2. 5 h) § PC lab sessions Robrecht Haesevoets (Robrecht. Haesevoets@cs. kuleuven. be)

MAS course material n transparencies – on-line n book/chapter/paper material ¨ M. Wooldridge, “An Introduction to Multi. Agent Systems” (Wiley, 2002) ¨ ¨ ¨ detailed description background information exercises examples. . . C. Larman, Applying UML and Patterns, chapter 33 M. Wooldridge, Reasoning about Rational Agents, chapter 1 -2 Bratman, Israel, Pollack, Plans and Resource-bounded Practical Reasoning Bonabeau, Dorigo, Theraulaz, Swarm Intelligence – From Natural to Artificial Systems, chapter 1 -2 Wooldridge, Jennings, Software Engineering with Agents: Pitfalls and Pratfalls …

MAS examination: project ¨ MAS project ¨ n n ¨ distribution of project description: April group of 2 students write a comprehensive document (10 -12 pages) prepare a short presentation / questions exam: discuss n ¨ “engineer a particular application using MAS” § analysis / architecture / design § make an implementation § report your paper Deadline n n n Wednesday, June 6 16 h 00 submit at secretariat Computer Science and on-line time mgt … !!

MAS course website n “Toledo” / Blackboard n n n http: //toledo. kuleuven. ac. be/ course material course ID: B-KUL-H 02 H 4 A discussion forum interesting links project information questions / remarks / suggestions /. . . register !!

MAS course overview n n 1. Introduction 2. Multi-Agent Systems n n general / application domains / examples / … 3. Multi-Agent Systems in Software Engineering n n (tentative) position of MAS in UP 4. Agent Architectures n agent architectures deductive reasoning agents / practical reasoning agents / reactive agents / hybrid agents n n 5. Communication n multi-agent architectures agent communication languages (ACLs) FIPA 6. Actions / Interactions / Cooperation n action models interaction cooperation n 7. Cooperative Distributed Problem Solving n 8. Ant- / Swarm-based Systems n n n 9. Coordination and Control Applications n n emergence ACO delegate MAS manufacturing control warehouse management 10. Conclusion n n Agent-Oriented Software Engineering methodologies pitfalls and pratfalls