Abstract
Cooperative tasks in Multi-Agent Reinforcement Learning (MARL) require agents to collectively maximize a shared return. Under the Centralized Training with Decentralized Execution (CTDE) paradigm, policy gradients have remained difficult to compute directly. Prior methods largely follow two approaches: independent factorized updates with centralized critics, which lack general joint-improvement guarantees without value decomposition assumptions, or alternating best-response updates, which can converge to suboptimal Nash Equilibria.
In this paper, we show the joint policy gradient admits an exact decentralized decomposition of per-agent terms, each formed from per-agent score functions and decentralized critics. Based on this decomposition, we develop Agent-Chained Policy Optimization (ACPO), where actors are trained independently, with their updates together constituting a single step on the joint policy gradient. Central to this result is a serialized view of the simultaneous joint decision in which agents commit actions one at a time, each conditioning on a belief over preceding actions that ties the independent per-agent updates into a single joint step.
We evaluate on-policy and off-policy instantiations of ACPO on Multi-Robot Warehouse, SMACv2, and MA-MuJoCo, where it outperforms strong baselines, with the gap widening as the number of agents grows.
Blogger's Review: The ACPO method addresses key challenges in Multi-Agent Reinforcement Learning through effective policy gradient decomposition, showcasing the potential for improved performance in complex cooperative scenarios. This innovation not only advances theoretical development but also provides fresh insights for practical applications. Particularly in multi-agent systems, the independent training strategy of ACPO opens up vast opportunities for future research.