Modify ActionInfo for Agent
2 次查看(过去 30 天)
显示 更早的评论
For reinforcement learning, I have defined my own environment class. Importantly, the ActionInfo of the environment is updated after every iteration based on a newly calculated component of the state, similar to the following:
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/771633/image.jpeg)
The environment works well enough. For the agent, I am using rlQAgent. However, the agent's ActionInfo is not getting updated after every iteration like the environment's ActionInfo, leading to a mismatch in the action space and thus an error. How would I modify the agent such that its action space gets updated along with the environment's after every iteration? I would like to do almost exactly as the above, except it would be done in the agent instead of the environment. I know that we can create custom agents, but I don't really need the agent to be customized apart from updating the action space after each iteration, and so I would like to avoid having to create a fully custom agent merely to alter one behavior of a predefined agent.
0 个评论
回答(0 个)
另请参阅
类别
在 Help Center 和 File Exchange 中查找有关 Environments 的更多信息
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!