Investigation of Independent Reinforcement Learning Algorithms in Multi-Agent Environments

TL;DR

This paper empirically compares independent reinforcement learning algorithms across various multi-agent environments, revealing their strengths and limitations in cooperative, competitive, and mixed settings, and highlighting the impact of recurrence in partially observable environments.

Contribution

It provides a comprehensive empirical analysis of independent RL algorithms in diverse multi-agent scenarios, filling a gap in understanding their practical performance.

Findings

Independent algorithms perform comparably to multi-agent algorithms in fully observable cooperative and competitive environments.

In mixed environments, independent algorithms excel individually but struggle with cooperation and competition.

Recurrence enhances learning in cooperative partially observable environments.

Abstract

Independent reinforcement learning algorithms have no theoretical guarantees for finding the best policy in multi-agent settings. However, in practice, prior works have reported good performance with independent algorithms in some domains and bad performance in others. Moreover, a comprehensive study of the strengths and weaknesses of independent algorithms is lacking in the literature. In this paper, we carry out an empirical comparison of the performance of independent algorithms on four PettingZoo environments that span the three main categories of multi-agent environments, i.e., cooperative, competitive, and mixed. We show that in fully-observable environments, independent algorithms can perform on par with multi-agent algorithms in cooperative and competitive settings. For the mixed environments, we show that agents trained via independent algorithms learn to perform well individually, but fail to learn to cooperate with allies and compete with enemies. We also show that adding recurrence improves the learning of independent algorithms in cooperative partially observable environments.