Mihai Barbuceanu, Wai-Kau Lo
In this paper we articulate a multi-level view of agent coordination and provide solutions for an integrated agent architecture that addresses all the levels. At the individual agent level, we model the decision making problem faced by individual agents that need to discover their highest utility goals and the plans to achieve them. Individual level plans normally contain goals that lie outside the agent's control. To achieve them, the agent needs to team up with other agents in the organization. At the organizational level, we show how organizational structures can be used to form the minimum cost teams needed to achieve such goals. Individual and organizational reasoning rely on knowing the utilities of the options available to agents. Often, these utilities are not given in advance, they must be discovered dynamically by market driven interaction. At the market level, we give a constraint optimization formulation to Multi Attribute Utility Theory and introduce interaction processes that allow agents to discover how to cooperate to optimize their objectives. All levels translate their specific models into a common reasoning infrastructure integrating randomized and systematic search.
Keywords: Multi-Agent Systems, Autonomous Agents
Citation: Mihai Barbuceanu, Wai-Kau Lo: Integrating Individual, Organizational and Market Level Reasoning for Agent Coordination. In W.Horn (ed.): ECAI2000, Proceedings of the 14th European Conference on Artificial Intelligence, IOS Press, Amsterdam, 2000, pp.343-347.