An Efficient Approach to Do Multi-agent Planning

Multi-agent planning is a difficult, yet under investigated class of planning problems. In the cooperative assumption, agents plan to achieve their individual goals independently, communicate with each other by continually sending and receiving messages from other agents. However, finding a globally consistent plan requires agents to interact with each other, which is time-consuming. In this paper, we describe a general framework that extracts the graph structure from causal constraints between agents in solving multi-agent planning problems to reduce the effort of repeatedly searching agents' consistent local plans and we theoretical analytically show that our approach can efficiently solve multi-agent planning problems with very complicated coupling level.