Dealing With Groups of Actions in Multiagent Markov Decision Processes

Guillaume Debras, Abdel-Illah Mouaddib, Laurent Jean Pierre, Simon Le Gloannec

2016

Abstract

Multiagent Markov Decision Processes (MMDPs) provide a useful framework for multiagent decision making. Finding solutions to large-scale problems or with a large number of agents however, has been proven to be computationally hard. In this paper, we adapt H-(PO)MDPs to multi-agent settings by proposing a new approach using action groups to decompose an initial MMDP into a set of dependent Sub-MMDPs where each action group is assigned a corresponding Sub-MMDP. Sub-MMDPs are then solved using a parallel Bellman backup to derive local policies which are synchronized by propagating local results and updating the value functions locally and globally to take the dependencies into account. This decomposition allows, for example, specific aggregation for each sub-MMDP, which we adapt by using a novel value function update. Experimental evaluations have been developed and applied to real robotic platforms showing promising results and validating our techniques.

Paper Citation

in Harvard Style

Debras G., Mouaddib A., Jean Pierre L. and Le Gloannec S. (2016). Dealing With Groups of Actions in Multiagent Markov Decision Processes . In Proceedings of the 8th International Joint Conference on Computational Intelligence - Volume 1: ECTA, (IJCCI 2016) ISBN 978-989-758-201-1, pages 49-58. DOI: 10.5220/0006048000490058

in Bibtex Style

@conference{ecta16,
author={Guillaume Debras and Abdel-Illah Mouaddib and Laurent Jean Pierre and Simon Le Gloannec},
title={Dealing With Groups of Actions in Multiagent Markov Decision Processes},
booktitle={Proceedings of the 8th International Joint Conference on Computational Intelligence - Volume 1: ECTA, (IJCCI 2016)},
year={2016},
pages={49-58},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0006048000490058},
isbn={978-989-758-201-1},
}

in EndNote Style

TY - CONF
JO - Proceedings of the 8th International Joint Conference on Computational Intelligence - Volume 1: ECTA, (IJCCI 2016)
TI - Dealing With Groups of Actions in Multiagent Markov Decision Processes
SN - 978-989-758-201-1
AU - Debras G.
AU - Mouaddib A.
AU - Jean Pierre L.
AU - Le Gloannec S.
PY - 2016
SP - 49
EP - 58
DO - 10.5220/0006048000490058