Expectation Maximization for Average Reward Decentralized POMDPs

摘要

Planning for multiple agents under uncertainty is often based on decentralized partially observable Markov decision processes (Dec-POMDPs), but current methods must de-emphasize long-term effects of actions by a discount factor. In tasks like wireless networking, agents are evaluated by average performance over time, both short and long-term effects of actions are crucial, and discounting based solutions can perform poorly. We show that under a common set of conditions expectation maximization (EM) for average reward Dec-POMDPs is stuck in a local optimum. We introduce a new average reward EM method; it outperforms a state of the art discounted-reward Dec-POMDP method in experiments.