Replace PI Controller with RL Agent for simple Transfer Function

Hello,

this question is related to my last question, which couldn't be answer yet.

I am not able to solve the simple task of replacing a PI Controller with DQN Agent, the training isn't going as planned.

I used the same system like the example:

https://de.mathworks.com/help/reinforcement-learning/ug/water-tank-reinforcement-learning-environment-model.html

But this still didn't work. Do you have any suggestions?

Answer 1

[Emmanouil Tzorakoleftherakis]

Please see answer here:

https://www.mathworks.com/matlabcentral/answers/779177-ddpg-agent-isn-t-learning-reward-0-for-every-episode?s_tid=srchtitle

[Sadagopan]

Hi,

This cannot be accepted answer beacuse i developed a simulink model based on differential equations of my system and the rl agent was learning and training properly, but when I developed the transfer function model for same environment with the input and output data and replaced my diff. eqn. system with the transfer function model, the training does not happen or the agent is not learning.