Sample-Based Policy Iteration for Constrained DEC-POMDPs

We introduce constrained DEC-POMDPs — an extension of the standard DEC-POMDPs that includes constraints on the optimality of the overall team rewards. Constrained DEC-POMDPs present a natural framework for modeling cooperative multi-agent problems with limited resources. To solve such DEC-POMDPs, we propose a novel sample-based policy iteration algorithm. The algorithm builds on multi-agent dynamic programming and benefits from several recent advances in DEC-POMDP algorithms such as MBDP [12] and TBDP [13]. Specifically, it improves the joint policy by solving a series of standard nonlinear programs (NLPs), thereby building on recent advances in NLP solvers. Our experimental results confirm the algorithm can efficiently solve constrained DECPOMDPs that cause general DEC-POMDP algorithms to fail.