Learning Cooperative Multi-Agent Policies with Partial Reward Decoupling

TL;DR

This paper introduces partial reward decoupling (PRD), a method that decomposes multi-agent RL problems into subproblems to improve credit assignment, data efficiency, and stability in cooperative settings.

Contribution

The paper proposes PRD, a novel approach that simplifies credit assignment in multi-agent RL by decomposing problems, leading to better performance than existing methods like COMA.

Findings

PRD reduces variance in policy gradient estimates.

PRD improves data efficiency and learning stability.

PRD outperforms COMA in various multi-agent tasks.

Abstract

One of the preeminent obstacles to scaling multi-agent reinforcement learning to large numbers of agents is assigning credit to individual agents' actions. In this paper, we address this credit assignment problem with an approach that we call \textit{partial reward decoupling} (PRD), which attempts to decompose large cooperative multi-agent RL problems into decoupled subproblems involving subsets of agents, thereby simplifying credit assignment. We empirically demonstrate that decomposing the RL problem using PRD in an actor-critic algorithm results in lower variance policy gradient estimates, which improves data efficiency, learning stability, and asymptotic performance across a wide array of multi-agent RL tasks, compared to various other actor-critic approaches. Additionally, we relate our approach to counterfactual multi-agent policy gradient (COMA), a state-of-the-art MARL algorithm, and empirically show that our approach outperforms COMA by making better use of information in agents' reward streams, and by enabling recent advances in advantage estimation to be used.