Coordinated Proximal Policy Optimization

TL;DR

CoPPO extends PPO to multi-agent systems by coordinating step sizes during policy updates, ensuring monotonic improvement, dynamic credit assignment, and outperforming baselines in cooperative tasks.

Contribution

The paper introduces CoPPO, a novel multi-agent extension of PPO that coordinates policy updates, guarantees monotonic improvement, and enhances performance in cooperative environments.

Findings

CoPPO outperforms strong baselines in cooperative matrix games.

CoPPO is competitive with MAPPO in StarCraft II tasks.

The method achieves dynamic credit assignment among agents.

Abstract

We present Coordinated Proximal Policy Optimization (CoPPO), an algorithm that extends the original Proximal Policy Optimization (PPO) to the multi-agent setting. The key idea lies in the coordinated adaptation of step size during the policy update process among multiple agents. We prove the monotonicity of policy improvement when optimizing a theoretically-grounded joint objective, and derive a simplified optimization objective based on a set of approximations. We then interpret that such an objective in CoPPO can achieve dynamic credit assignment among agents, thereby alleviating the high variance issue during the concurrent update of agent policies. Finally, we demonstrate that CoPPO outperforms several strong baselines and is competitive with the latest multi-agent PPO method (i.e. MAPPO) under typical multi-agent settings, including cooperative matrix games and the StarCraft II micromanagement tasks.