Parallel Training of Multiple RL Agents in same environment

Question

Federico Toso 2024-2-18

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2083473-parallel-training-of-multiple-rl-agents-in-same-environment

评论： Federico Toso 2024-2-23

In the context of Reinforcement Learning Toolbox, it is possible to set "UseParallel" to "true" within "rlTrainingOptions" in order to run multiple simulations in parallel and improve the training speed for the agent.

However, if my goal is to train multiple agents in same environment I have to use "rlMultiAgentTrainingOptions" in order to specify my options. According to the documentation, this object does not have a "UseParallel" property. Does this mean that parallel training is currently not supported for multi-agent scenario? If this is the case, is there a known workaround for this?

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Emmanouil Tzorakoleftherakis 2024-2-20

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2083473-parallel-training-of-multiple-rl-agents-in-same-environment#answer_1412908

Parallel training is currently not supported for multi-agent reinforcement learning. One thing you could do is train the agents sequentially in parallel, but this will only work if the agents are "decoupled", i.e. they operate on different components on the environment and can thus be trained separately.