Skip to content

Latest commit

 

History

History
24 lines (10 loc) · 910 Bytes

File metadata and controls

24 lines (10 loc) · 910 Bytes

多智能体强化学习论文

===

Is Cooperative MARL Solved?

个人认为,cooperative multi-agent reinforcement learning在发现了基于策略梯度的sequential rollout机制以后就算是解决了,剩下就是对sequential rollout工程实现上的修补改进(比如引入order matters, transformer结构, distill方法等)。

多智能体Transfermor

多智能体在连续动作空间中使用Qmix方法(FACMAC)