Tree-based pruning for multiagent POMDPs with delayed communication

Frans A. Oliehoek, Matthijs T. J. Spaan

Abstract: Multiagent POMDPs provide a powerful framework for optimal decision making under the assumption of instantaneous communication. We focus on a delayed communication setting (MPOMDP-DC), in which broadcast information is delayed by at most one time step. Such an assumption is in fact more appropriate for applications in which response time is critical. However, naive application of incremental pruning, the core of many state-of-the-art POMDP techniques, is intractable for MPOMDP-DCs. We overcome this problem by introducing a tree-based pruning technique. Experiments show that the method outperforms naive incremental pruning by orders of magnitude, allowing for the solution of larger problems.