Assigning Credit with Partial Reward Decoupling in Multi-Agent Proximal Policy Optimization
This addresses the credit assignment problem in multi-agent reinforcement learning, which is crucial for scaling to large teams, though it is incremental as it builds on existing MAPPO methods.
The paper tackles the credit assignment problem in multi-agent proximal policy optimization (MAPPO) by proposing PRD-MAPPO, which uses partial reward decoupling to dynamically form subgroups and streamline credit assignment, resulting in significantly higher data efficiency and asymptotic performance compared to MAPPO and other state-of-the-art methods in tasks like StarCraft II.
Multi-agent proximal policy optimization (MAPPO) has recently demonstrated state-of-the-art performance on challenging multi-agent reinforcement learning tasks. However, MAPPO still struggles with the credit assignment problem, wherein the sheer difficulty in ascribing credit to individual agents' actions scales poorly with team size. In this paper, we propose a multi-agent reinforcement learning algorithm that adapts recent developments in credit assignment to improve upon MAPPO. Our approach leverages partial reward decoupling (PRD), which uses a learned attention mechanism to estimate which of a particular agent's teammates are relevant to its learning updates. We use this estimate to dynamically decompose large groups of agents into smaller, more manageable subgroups. We empirically demonstrate that our approach, PRD-MAPPO, decouples agents from teammates that do not influence their expected future reward, thereby streamlining credit assignment. We additionally show that PRD-MAPPO yields significantly higher data efficiency and asymptotic performance compared to both MAPPO and other state-of-the-art methods across several multi-agent tasks, including StarCraft II. Finally, we propose a version of PRD-MAPPO that is applicable to \textit{shared} reward settings, where PRD was previously not applicable, and empirically show that this also leads to performance improvements over MAPPO.