Selectively Sharing Experiences Improves Multi-Agent Reinforcement Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Gerstgrasser, Matthias, Danino, Tom, Keren, Sarah
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909179564785664
author Gerstgrasser, Matthias
Danino, Tom
Keren, Sarah
author_facet Gerstgrasser, Matthias
Danino, Tom
Keren, Sarah
contents 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. A reference implementation of our algorithm is available at https://github.com/mgerstgrasser/super.
format Preprint
id arxiv_https___arxiv_org_abs_2311_00865
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Selectively Sharing Experiences Improves Multi-Agent Reinforcement Learning
Gerstgrasser, Matthias
Danino, Tom
Keren, Sarah
Machine Learning
Artificial Intelligence
Multiagent Systems
Robotics
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. A reference implementation of our algorithm is available at https://github.com/mgerstgrasser/super.
title Selectively Sharing Experiences Improves Multi-Agent Reinforcement Learning
topic Machine Learning
Artificial Intelligence
Multiagent Systems
Robotics
url https://arxiv.org/abs/2311.00865