Conflict Resolution in Probabilistic Multi-Agent Systems

May 2002

Proceedings of the Fifteenth International Florida Artificial Intelligence Research Society Conference (FLAIRS 2002)

Volume

Proceedings of the Fifteenth International Florida Artificial Intelligence Research Society Conference (FLAIRS 2002)

All Papers

Knowledge representation in the Multi-agent systems (MAS) can be characterized by Probabilistic Network. Coordination among probabilistic agents can be achieved on account of their common knowledge, while agent can obtain his own knowledge locally and variously. To construct the common knowledge for the MAS, the conflict among local knowledge of all agents has to be identified and resolved. We propose the necessary and sufficient conditions for their knowledge structure. We also address that the redundancy has to be eliminated at the beginning in case of misleading conclusion. An algorithm is suggested to construct the structure of common knowledge in terms of Bayesian DAG.