Celebrating Diversity in Shared Multi-Agent Reinforcement Learning

TL;DR

This paper introduces a novel approach to enhance diversity in shared multi-agent reinforcement learning by combining information-theoretical regularization with agent-specific modules, leading to improved coordination and state-of-the-art results.

Contribution

It presents a new method that promotes diversity in shared MARL through mutual information regularization and agent-specific modules, advancing the field's ability to handle complex cooperative tasks.

Findings

Achieves state-of-the-art performance on Google Research Football.

Outperforms existing methods on StarCraft II micromanagement tasks.

Enhances agent diversity and coordination in shared MARL settings.

Abstract

Recently, deep multi-agent reinforcement learning (MARL) has shown the promise to solve complex cooperative tasks. Its success is partly because of parameter sharing among agents. However, such sharing may lead agents to behave similarly and limit their coordination capacity. In this paper, we aim to introduce diversity in both optimization and representation of shared multi-agent reinforcement learning. Specifically, we propose an information-theoretical regularization to maximize the mutual information between agents' identities and their trajectories, encouraging extensive exploration and diverse individualized behaviors. In representation, we incorporate agent-specific modules in the shared neural network architecture, which are regularized by L1-norm to promote learning sharing among agents while keeping necessary diversity. Empirical results show that our method achieves state-of-the-art performance on Google Research Football and super hard StarCraft II micromanagement tasks.