Bi-level Mean Field: Dynamic Grouping for Large-Scale MARL

TL;DR

This paper introduces a Bi-level Mean Field approach with dynamic agent grouping using VAE to improve large-scale MARL, effectively capturing agent diversity and reducing aggregation noise for better learning performance.

Contribution

It proposes a novel Bi-level Mean Field method with dynamic grouping via VAE, addressing aggregation noise and agent heterogeneity in large-scale MARL.

Findings

BMF outperforms state-of-the-art methods across various tasks.

Dynamic grouping reduces aggregation noise effectively.

Bi-level interaction captures both inter- and intra-group dynamics.

Abstract

Large-scale Multi-Agent Reinforcement Learning (MARL) often suffers from the curse of dimensionality, as the exponential growth in agent interactions significantly increases computational complexity and impedes learning efficiency. To mitigate this, existing efforts that rely on Mean Field (MF) simplify the interaction landscape by approximating neighboring agents as a single mean agent, thus reducing overall complexity to pairwise interactions. However, these MF methods inevitably fail to account for individual differences, leading to aggregation noise caused by inaccurate iterative updates during MF learning. In this paper, we propose a Bi-level Mean Field (BMF) method to capture agent diversity with dynamic grouping in large-scale MARL, which can alleviate aggregation noise via bi-level interaction. Specifically, BMF introduces a dynamic group assignment module, which employs a Variational AutoEncoder (VAE) to learn the representations of agents, facilitating their dynamic grouping over time. Furthermore, we propose a bi-level interaction module to model both inter- and intra-group interactions for effective neighboring aggregation. Experiments across various tasks demonstrate that the proposed BMF yields results superior to the state-of-the-art methods.