Consensus Learning for Cooperative Multi-Agent Reinforcement Learning

TL;DR

This paper introduces a consensus learning approach for cooperative multi-agent reinforcement learning, enabling agents to infer shared consensus from local observations to improve cooperation during decentralized execution.

Contribution

It proposes a novel consensus inference method inspired by viewpoint invariance and contrastive learning, enhancing cooperation in multi-agent systems with minimal model modifications.

Findings

Achieved convincing results on fully cooperative tasks.

Method extends easily to various existing algorithms.

Improved agent cooperation during decentralized execution.

Abstract

Almost all multi-agent reinforcement learning algorithms without communication follow the principle of centralized training with decentralized execution. During centralized training, agents can be guided by the same signals, such as the global state. During decentralized execution, however, agents lack the shared signal. Inspired by viewpoint invariance and contrastive learning, we propose consensus learning for cooperative multi-agent reinforcement learning in this paper. Although based on local observations, different agents can infer the same consensus in discrete space. During decentralized execution, we feed the inferred consensus as an explicit input to the network of agents, thereby developing their spirit of cooperation. Our proposed method can be extended to various multi-agent reinforcement learning algorithms with small model changes. Moreover, we carry out them on some fully cooperative tasks and get convincing results.