Course Modules

Module 1: Introduction to Distributed Problem Solving/Multi-agent Systems

• Features of such systems: loose coupling, decentralized control and information, asynchronous, cooperation/competition (subproblem interaction), limited communication bandwidth, likelihood of hardware failure and agent heterogeneity, and implications of these features on uncertainty, incompleteness and errorfulness of both domain and control problem solving.
• Application areas: distributed interpretation, distributed planning and control, and cooperative interaction among knowledge based system/people. Examples of applications.
• Key dimensions in judging different system architectures: numbers of agents, heterogeneity among agents, sophistication of local agent problem solving, type of information communicated, and coordination mechanisms.
• Long-term research goals of field and open problems.

Module 2: Coordination

• What is it and why is it difficult?
• Distributed Search and the issues of control, solution and cooperative control uncertainty. Issues of distraction and poaching. The need for meta-level information and the applicability of meta-level control.
• Relationship between agent problem-solving and coordination.
• Coordination relationships and the importance of their quantitative characteristic.  Discussion of TAEMS as a modeling framework. Relationship to distributed scheduling. Issue of predictability vs. responsiveness
• Generic partial global planning. Sophisticated local control and its relationship to coordination issues.
• Other approaches to Coordination: Use of Belief-desire-intention semantics (Jennings, Tambe, Grosz and Georgeff); KQML performas, Reactive Control.

Module 3: Negotiation

• Definition and its importance, the different phases of distributed problem solving to which it can be applied, a generic model of the negotiation process, concepts of multi-linked and multi-phased negotiation
• Issues in automated contracting
• Tolerance of inconsistency, functionally/accurate/cooperative problem solving (DRESUN), exploiting subproblem interactions, negotiated search (Moehlman, Lander), Persuader (Sycara), distributed ATMS(Huhns), Traconet (Sandholm),  levelled commitment protocols (Sandholm), DARM (Neiman), TEAM (Lander).

• Different forms of problem-solving organizations, static organization forms, organizational self-design, the need for problem-solving roles, organizational knowledge in terms of who, what, when, where, how and why.
• Organization as a framework for mediating between subjective and objective reality, and contingency theory.
• Hierarchical control via modulating local control, skeptical nodes, social laws( Tanenbaum), action-at-a-distance(Gasser)
• Detailed discussion of Durfee/Montgomery on organizational design, Ishida/Gasser, Decker on load balancing

• Logics of knowledge and belief, possible world semantics, agent oriented programming (Singh, Georgeff, Grosz, Shoham)
• Speech acts and their use as a framework for agent communication
• Game and decision theoretic models for coordination and negotiation among competitive/hostile agents (Rosenschein)
• Market Mechanisms (Wellman, Huberman)

Module 6: Role of Learning

The role of reinforcement learning and explanation-based learning for use in organization self-design, coordination rules, and social laws.