Selectively Sharing Experiences Improves Multi-Agent Reinforcement Learning

Abstract

We present a novel multi-agent RL approach, Selective Multi-Agent Prioritized Experience Relay, in which agents share with other agents a limited number of transitions they observe during training. The intuition behind this is that even a small number of relevant experiences from other agents could help each agent learn. Unlike many other multi-agent RL algorithms, this approach allows for largely decentralized training, requiring only a limited communication channel between agents. We show that our approach outperforms baseline no-sharing decentralized training and state-of-the art multi-agent RL algorithms. Further, sharing only a small number of highly relevant experiences outperforms sharing all experiences between agents, and the performance uplift from selective experience sharing is robust across a range of hyperparameters and DQN variants.

Cite

Text

Gerstgrasser et al. "Selectively Sharing Experiences Improves Multi-Agent Reinforcement Learning." Neural Information Processing Systems, 2023.

Markdown

[Gerstgrasser et al. "Selectively Sharing Experiences Improves Multi-Agent Reinforcement Learning." Neural Information Processing Systems, 2023.](https://mlanthology.org/neurips/2023/gerstgrasser2023neurips-selectively/)

BibTeX

@inproceedings{gerstgrasser2023neurips-selectively,
  title     = {{Selectively Sharing Experiences Improves Multi-Agent Reinforcement Learning}},
  author    = {Gerstgrasser, Matthias and Danino, Tom and Keren, Sarah},
  booktitle = {Neural Information Processing Systems},
  year      = {2023},
  url       = {https://mlanthology.org/neurips/2023/gerstgrasser2023neurips-selectively/}
}