Monte-Carlo Expectation Maximization for Decentralized POMDPs
Abstract
We address two significant drawbacks of state-of-the-art solvers of decentralized POMDPs (DEC-POMDPs): the reliance on complete knowledge of the model and limited scalability as the complex-ity of the domain grows. We extend a recently proposed approach for solving DEC-POMDPs vi-a a reduction to the maximum likelihood problem, which in turn can be solved using EM. We intro-duce a model-free version of this approach that em-ploys Monte-Carlo EM (MCEM). While a naı̈ve implementation of MCEM is inadequate in multi-agent settings, we introduce several improvements in sampling that produce high-quality results on a variety of DEC-POMDP benchmarks, including large problems with thousands of agents. 1
Cite
Text
Wu et al. "Monte-Carlo Expectation Maximization for Decentralized POMDPs." International Joint Conference on Artificial Intelligence, 2013.Markdown
[Wu et al. "Monte-Carlo Expectation Maximization for Decentralized POMDPs." International Joint Conference on Artificial Intelligence, 2013.](https://mlanthology.org/ijcai/2013/wu2013ijcai-monte/)BibTeX
@inproceedings{wu2013ijcai-monte,
title = {{Monte-Carlo Expectation Maximization for Decentralized POMDPs}},
author = {Wu, Feng and Zilberstein, Shlomo and Jennings, Nicholas R.},
booktitle = {International Joint Conference on Artificial Intelligence},
year = {2013},
pages = {397-403},
url = {https://mlanthology.org/ijcai/2013/wu2013ijcai-monte/}
}