The Surprising Effectiveness of PPO in Cooperative Multi-Agent GamesDownload PDF

04 Jun 2022, 06:09 (modified: 09 Oct 2022, 22:33)NeurIPS 2022 Datasets and Benchmarks Readers: Everyone
Keywords: Multi-Agent Reinforcement Learning, Proximal Policy Optimization, Cooperative Games
TL;DR: We demonstrate PPO's effectiveness in popular multi-agent benchmarks and analyze its properties and implementation details through empirical studies.
Abstract: Proximal Policy Optimization (PPO) is a ubiquitous on-policy reinforcement learning algorithm but is significantly less utilized than off-policy learning algorithms in multi-agent settings. This is often due to the belief that PPO is significantly less sample efficient than off-policy methods in multi-agent systems. In this work, we carefully study the performance of PPO in cooperative multi-agent settings. We show that PPO-based multi-agent algorithms achieve surprisingly strong performance in four popular multi-agent testbeds: the particle-world environments, the StarCraft multi-agent challenge, the Hanabi challenge, and Google Research Football, with minimal hyperparameter tuning and without any domain-specific algorithmic modifications or architectures. Importantly, compared to competitive off-policy methods, PPO often achieves competitive or superior results in both final returns and sample efficiency. Finally, through ablation studies, we analyze implementation and hyperparameter factors that are critical to PPO's empirical performance, and give concrete practical suggestions regarding these factors. Our results show that when using these practices, simple PPO-based methods are a strong baseline in cooperative multi-agent reinforcement learning. Source code is released at
Supplementary Material: pdf
License: MIT License
Author Statement: Yes
Contribution Process Agreement: Yes
In Person Attendance: Yes
30 Replies