How the generated parameters of agent are related PI/PID controller in reinforcement learning algorithm

Respected all,

The following example shows the PI controller design using TD3 reinforcement learning algorithm. How the agent parameters are related to PI controller gains for the following example.

Tune PI Controller using Reinforcement Learning - MATLAB & Simulink - MathWorks India

Answer 1

[Emmanouil Tzorakoleftherakis]

This example uses an actor network that works as a PI controller. This is possible because we are using a customer layer called fullyConnectedPILayer.

After training, the weights of the fullyConnectedPILayer are effectively the weights of the tuned PI controller.

Hope that helps

[Kurnam Gnaneshwar]

Dear sir,

I agree to the answer. However, it is not understandable to the reader.

Can you please provide editable actor model which is designed uisng fullyConnectedPILayer. Then, we can modify according to the our needs.

Thank You

[Emmanouil Tzorakoleftherakis]

Not sure I fully understand. The actor model in the example is fully editable