How to set the state with different variables in properties?

Question

Yang Chen 2023-3-5

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1923440-how-to-set-the-state-with-different-variables-in-properties

回答： Emmanouil Tzorakoleftherakis 2023-3-6

I am writing the custom reinforcement learning environment via template. I want to initialize state within properties. The following is from my current scripts. The scripts are based on the class called rl.env.MATLABEnvironment. The properties of this class can't accept the structure definition.

properties

% Initialize system state [schedule,ppath,completionTime,computeDuring]'

% schedule = [];

% ppath = [];

% completionTime = Inf;

% computeDuring = 0;

State = zeros(4,1)

end

However, I don't know how to definite State. In the above case, State consists of schedule, ppath, completionTime and computerDuring, where schedule and ppath are vectors, while completionTime and computerDuring are numbers. If I just define them as zeros(4,1), it doesn't work.

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 2023-3-6

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1923440-how-to-set-the-state-with-different-variables-in-properties#answer_1186540

在 MATLAB Online 中打开

Hi Yang,

We have an example in Reinforcement Learning Toolbox that does training based on nonhomogeneous observations, and specifically a combination of scalars and images. This example is available as a predefined environment and you can actually check the code to see how we implemented it. If you do

edit rl.env.AbstractSimplePendlumWithImage

you will see that we create the state in the properties similar to what you do but this is only for the scalar states of the dynamical equations. The nonhomogeneous observations are passed internally as cell arrays starting from the 'reset' function.

Hope that helps