Optimization-based tuning–PID controller with Delay

[caxelrud]

I tried the example of optimized-based tuning PID from the documentation at:
https://docs.sciml.ai/ControlSystems/stable/examples/automatic_differentiation/
I modified the process transfer function to one that includes dead-time as: P = tf(1, [4, 1.0])*delay(1) .
I got the following error when try to execute: res = solve(prob, solver).
ERROR: The advanced interface to the function feedback (with connection keyword arguments) is currently not supported for LFT systems (such as nonlinear and time-delay systems)

From the documentation, at Properties of delay systems, I see examples of the feedback function with process with time-delay.
Let me know if it is a bug or a current limitation.
Thanks!

[baggepinnen]

You can create feedback connections with delay systems, but not using the advanced version of the feedack function that is used inside the extended_gangoffour function, this is a current limitation. You could reformulate the problem to work on discrete-time systems instead, and discretize the delay exactly. To do this, call c2d on your system model before the optimization, pass the same sample time Ts to pid in the systemspid function. After that, all computations should be performed on a non-delayed discrete-time model with internal state representing the delay instead

[caxelrud]

Thanks baggepinnen!
It worked with the instructions you recommended.

[baggepinnen]

Great!