Semantics of multiply sectioned bayesian networks for cooperative multi-agent distributed interpretation

摘要

In order to represent cooperative multi-agents who must reason with uncertain knowledge, a coherent, framework is necessary. We choose multipley sectioned Bayesian networks (MSBNs) as the basis for this study because they are based on well established theory on Bayesian networks and because they are modular. In this paper, we focus on the semantics of a MSBN-based multi-agent system (MAS) for cooperative distributed interpretation. In particular, we establish the conditions under which the joint probability distribution of a MSBN-based MAS can be meaningfull interpreted. These conditions imply that a coherent MSBN-based MAS can be constructed using agents built by different developers. We show how the conditions can be satisfied technically under such a context.