starry-sky6688 / MADDPG

Pytorch implementation of the MARL algorithm, MADDPG, which correspondings to the paper "Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments".
544 stars 83 forks source link

为什么这里maddpg/maddpy.py里 其他agent的动作不变呢,王树森老师的《深度强化学习》书本里面说是要变的,请您解答一下,感谢~ #19

Closed zzhuncle closed 2 years ago

zzhuncle commented 2 years ago

image image

starry-sky6688 commented 2 years ago

你说的这个书上应该是改进过的,MADDPG的论文里,其他agent的动作是不变的,你看一下论文里的伪代码就知道了;

另外你还可以去看一下MAAC,应该也提到了MADDPG的这个问题

zzhuncle commented 2 years ago

谢谢了,我看到了,MAAC还用了自注意力、SAC、COMA,大杂烩,现在还在看MADDPG和MAAC,我感觉是不是过时了hhh,想问一下大佬最近有啥推荐的MARL的论文看一看~

starry-sky6688 commented 2 years ago

我也有段时间没看最新的MARL论文了,去看看NIPS和ICML里相关的吧,知乎搜,有人会把这些会议的强化学习论文列出来

zzhuncle commented 2 years ago

谢谢~