More Centralized Training, Still Decentralized Execution: Multi-Agent Conditional Policy Factorization
This work addresses a key bottleneck in MARL for partially observable environments, offering a method to improve policy learning without sacrificing decentralized execution, though it is incremental in nature.
The paper tackles the problem of learning optimal joint policies in cooperative multi-agent reinforcement learning by explicitly considering agent dependencies during centralized training, while still enabling decentralized execution. It proposes MACPF, which theoretically ensures factorization and empirically shows better performance or faster convergence than baselines in various tasks.
In cooperative multi-agent reinforcement learning (MARL), combining value decomposition with actor-critic enables agents to learn stochastic policies, which are more suitable for the partially observable environment. Given the goal of learning local policies that enable decentralized execution, agents are commonly assumed to be independent of each other, even in centralized training. However, such an assumption may prohibit agents from learning the optimal joint policy. To address this problem, we explicitly take the dependency among agents into centralized training. Although this leads to the optimal joint policy, it may not be factorized for decentralized execution. Nevertheless, we theoretically show that from such a joint policy, we can always derive another joint policy that achieves the same optimality but can be factorized for decentralized execution. To this end, we propose multi-agent conditional policy factorization (MACPF), which takes more centralized training but still enables decentralized execution. We empirically verify MACPF in various cooperative MARL tasks and demonstrate that MACPF achieves better performance or faster convergence than baselines. Our code is available at https://github.com/PKU-RL/FOP-DMAC-MACPF.