ColosseumRL: A Framework for Multiagent Reinforcement Learning in $N$-Player Games
Abstract
Much of recent success in multiagent reinforcement learning has been in two-player zero-sum games. In these games, algorithms such as fictitious self-play and minimax tree search can converge to an approximate Nash equilibrium. While playing a Nash equilibrium strategy in a two-player zero-sum game is optimal, in an -player general sum game, it becomes a much less informative solution concept. Despite the lack of a satisfying solution concept, -player games form the vast majority of real-world multiagent situations. In this paper we present a new framework for research in reinforcement learning in -player games. We hope that by analyzing behavior learned by agents in these environments the community can better understand this important research area and move toward meaningful solution concepts and research directions. The implementation and additional information about this framework can be found at https://colosseumrl.igb.uci.edu/.
Cite
@article{arxiv.1912.04451,
title = {ColosseumRL: A Framework for Multiagent Reinforcement Learning in $N$-Player Games},
author = {Alexander Shmakov and John Lanier and Stephen McAleer and Rohan Achar and Cristina Lopes and Pierre Baldi},
journal= {arXiv preprint arXiv:1912.04451},
year = {2019}
}
Comments
Accepted for the 2020 AAAI Spring Symposium, Challenges and Opportunities for Multi-Agent Reinforcement Learning. Source code available at https://github.com/colosseumrl/colosseumrl/