Context-Aware Bayesian Network Actor-Critic Methods for Cooperative Multi-Agent Reinforcement Learning

TL;DR

This paper introduces a Bayesian network approach to model correlated actions in cooperative multi-agent reinforcement learning, providing theoretical justification and practical algorithms that improve coordination and scalability.

Contribution

It proposes a novel Bayesian network framework for action correlation in MARL, with theoretical analysis and algorithms that adaptively learn context-aware policies.

Findings

Theoretical proof of convergence to Nash equilibria.

Empirical improvements on MARL benchmarks.

Dynamic adjustment of policy sparsity enhances performance.

Abstract

Executing actions in a correlated manner is a common strategy for human coordination that often leads to better cooperation, which is also potentially beneficial for cooperative multi-agent reinforcement learning (MARL). However, the recent success of MARL relies heavily on the convenient paradigm of purely decentralized execution, where there is no action correlation among agents for scalability considerations. In this work, we introduce a Bayesian network to inaugurate correlations between agents' action selections in their joint policy. Theoretically, we establish a theoretical justification for why action dependencies are beneficial by deriving the multi-agent policy gradient formula under such a Bayesian network joint policy and proving its global convergence to Nash equilibria under tabular softmax policy parameterization in cooperative Markov games. Further, by equipping existing MARL algorithms with a recent method of differentiable directed acyclic graphs (DAGs), we develop practical algorithms to learn the context-aware Bayesian network policies in scenarios with partial observability and various difficulty. We also dynamically decrease the sparsity of the learned DAG throughout the training process, which leads to weakly or even purely independent policies for decentralized execution. Empirical results on a range of MARL benchmarks show the benefits of our approach.