Multi-agent, Multi-target Path Planning in Markov Decision Processes

Missions for autonomous systems often require agents to visit multiple targets in complex operating conditions. This work considers the problem of visiting a set minimum time by team non-communicating Markov decision process (MDP). The single-agent is at least NP-complete reducing it Hamiltonian path problem. We first discuss an optimal algorithm based on Bellman's optimality equation that exponential number target states. Then, we trade-off complexity presenting suboptimal polynomial each step. prove proposed generates policies certain classes MDPs. Extending our procedure multi-agent case, propose partitioning approximately minimizes expected targets. partitions clustered scenarios. present performance algorithms random MDPs and gridworld environments inspired ocean dynamics. show are much faster than more currently available heuristic.