arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2306.02430cs.MAcs.LG
收藏

面向多智能体强化学习的分布式价值函数分解统一框架

A Unified Framework for Factorizing Distributional Value Functions for Multi-Agent Reinforcement Learning

National Tsing Hua University(国立清华大学);NVIDIA Corporation(英伟达公司)

Wei-Fang Sun, Cheng-Kuang Lee, Simon See, Chun-Yi Lee

AI总结:

针对完全协作 MARL 中部分可观测与策略非平稳导致的随机性,提出统一框架 DFAC,将分布式 RL 与价值函数分解结合,并在矩阵博弈及 SMAC 高难度地图上验证其性能优势。

摘要:

在完全协作式多智能体强化学习(MARL)设定中,由于每个智能体的部分可观测性以及其他智能体持续变化的策略,环境具有高度随机性。为解决上述问题,我们提出了一个名为 DFAC 的统一框架,用于将分布式强化学习与价值函数分解方法相结合。该框架推广了期望值价值函数分解方法,使其能够实现回报分布的分解。为验证 DFAC,我们首先展示了其在一个具有随机奖励的简单矩阵博弈中分解价值函数的能力。随后,我们在 StarCraft Multi-Agent Challenge 的所有 Super Hard 地图以及六个自行设计的 Ultra Hard 地图上进行实验,结果表明 DFAC 能够优于多个基线方法。

英文摘要:

In fully cooperative multi-agent reinforcement learning (MARL) settings, environments are highly stochastic due to the partial observability of each agent and the continuously changing policies of other agents. To address the above issues, we proposed a unified framework, called DFAC, for integrating distributional RL with value function factorization methods. This framework generalizes expected value function factorization methods to enable the factorization of return distributions. To validate DFAC, we first demonstrate its ability to factorize the value functions of a simple matrix game with stochastic rewards. Then, we perform experiments on all Super Hard maps of the StarCraft Multi-Agent Challenge and six self-designed Ultra Hard maps, showing that DFAC is able to outperform a number of baselines.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑