QueryPOMDP: POMDP-based Communication in Multiagent Systems⋆

Abstract. Decentralized Partially Observable Markov Decision Pro-cesses (Dec-POMDPs) provide powerful modeling tools for multiagent decision-making in the face of uncertainty, but solving these models comes at a very high computational cost. Two avenues for side-stepping the computational burden can be identified: structured interactions be-tween agents and intra-agent communication. In this paper, we focus on the interplay between these concepts, namely how sparse interactions impact the communication needs. A key insight is that in domains with local interactions the amount of communication necessary for successful joint behavior can be heavily reduced, due to the limited influence be-tween agents. We exploit this insight by deriving local POMDP models that optimize each agent’s communication behavior. Our experimental results show that our approach successfully exploits sparse interactions: we can effectively identify the situations in which it is beneficial to com-municate, as well as trade off the cost of communication with overall task performance