Planning in Multiagent Expedition with Collaborative Design Networks

DEC-POMDPs provide formal models of many cooperative multiagent problems, but their complexity is NEXP-complete in general. We investigate a sub-class of DEC-POMDPs termed multiagent expedition. A typical instance consists of an area populated by mobile agents. Agents have no prior knowledge of the area, have limited sensing and communication, and effects of their actions are uncertain. Success relies on planing actions that result in high accumulated rewards. We solve an instance of multiagent expedition based on collaborative design network, a decision theoretic multiagent graphical model. We present a number of techniques employed in knowledge representation and demonstrate the superior performance of our system in comparison to greedy agents experimentally.