How to run the simulink model when implementing custom RL training?

Question

Yihao Wan 2023-5-25

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1973424-how-to-run-the-simulink-model-when-implementing-custom-rl-training

评论： Emmanouil Tzorakoleftherakis 2023-5-25

Hello, I am developing a custom training of RL DQN agent based on the link, however, how should I adapt it to the simulink environment?

Especially for the code below, when applying an action to the environment, the step is not applicable for a simulink model. How should I solve this issue? Thanks in advance.

% Apply the action to the environment

% and obtain the resulting observation and reward.

[nextObs,reward,isdone] = step(env,action{1});

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 2023-5-25

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1973424-how-to-run-the-simulink-model-when-implementing-custom-rl-training#answer_1244699

The way to do it would be to use runEpisode

2 个评论
显示无隐藏无

Yihao Wan 2023-5-25

Thanks for the reply. The runEpisode would run all the steps rather than single step right? Then how should I modify the rest single step training iterations? I am refering to this example.

Thanks a lot.

Emmanouil Tzorakoleftherakis 2023-5-25

The example you are showing is model-based RL, it's different from what you mentioned at the beginning.

With runEpisode you have the flexibility of running the entire episode and learning after, or learning at every step. For that you can use the processExperiencefcn shown in the doc. This example shows how you can implement it.

请先登录，再进行评论。