Can I use well-trained ANN model and "place" function for Hardware in the loop (HIL) testing?

I have a trained ANN in MATLAB that estimates the plant’s parameters. These estimates go to an adaptive-control block that recomputes the controller gain K with the place function.

The snag: place is not supported for code generation.

Can I keep the ANN and adaptive-control logic (including the place call) running in MATLAB/Python on a host PC, while

• the plant runs in real-time on a simulator such as Typhoon HIL or OPAL-RT, and
• the main controller code executes on a TI C2000 DSP?

Hope someone can help me for this question!

Answer 1

[Manish]

Hi,

I understand that you want to generate code for the 'place' function.

This can be achieved with the help of 'coder.extrinsic'

Here is a sample code snippet:

function K = computeK(A, B, p)
coder.extrinsic('place');
K = zeros(size(B, 2), size(A, 1));
K(:,:) = place(A, B, p);
end

Refer to the documentation link below for a better understanding:

• https://www.mathworks.com/help/simulink/slref/coder.extrinsic.html

Hope it helps!

[Quang Manh]

I tried to use "coder.extrinsic" before however the simulation time is much more slower.

Now, I am considering the way to calculate controller gain K in Maltab environment so I don't need to generate code for "place" function. General idea is for example A value is sent from simulink model (while running) to MATLAB environment, then ANN using A to estimate B, then B is used to calculate adaptive controller gain K, then K is fed back to the Simulink model.

I think it's quite complicated.