Problem using lsqnonlin with a simulink model

Hello, I'm trying to use an lsqnonlin optimization with a simulink model: the simulink model requires as input the variables "sigma_1" and "Km_1", which are the parameters that I want to optimize. since the error function is obtained from the output of the model I inserted all in the function err_fun. The code I used is the following:

p = lsqnonlin(@(p) err_fun(p, acq_data),p0, lb, ub, options);

function err = err_fun(p, acq_data)
   sigma_1 = p(1);
   Km_1 = p(2);
   duration = 0.14;
   
   sim('Simulink_model',[0,duration]); 
   S_model = spostamento.signals.values;
   
   err = S_model - acq_data;
end

This code gives me an error, which I am almost sure is related to the fact that sigma_1 and Km_1 result undefined for the simulink model. The errore is the following:

Error using Launcher>err_fun (line 83)
Error due to multiple causes.

Error in Launcher>@(p)err_fun(p,acq_data)

Error in lsqnonlin (line 206)
            initVals.F = feval(funfcn{3},xCurrent,varargin{:});

Error in Launcher (line 76)
p = lsqnonlin(@(p) err_fun(p, acq_data),p0, lb, ub, options);
Caused by:
    Error using Launcher>err_fun (line 83)
    Error evaluating parameter 'stiff_up' in 'VESevo_sim_V1/Translational Hard Stop1'
        Error using Launcher>err_fun (line 83)
        Undefined function or variable 'Km_1'.
            Error using Launcher>err_fun (line 83)
            Variable 'Km_1' has been deleted from base workspace.
            Suggested Actions:
                • Undo the deletion. - Fix
                • Load a file into base workspace. - Fix
                • Create a new variable. - Fix
    Error using Launcher>err_fun (line 83)
    Error evaluating parameter 'D_up' in 'VESevo_sim_V1/Translational Hard Stop1'
        Error using Launcher>err_fun (line 83)
        Undefined function or variable 'sigma_1'.
            Error using Launcher>err_fun (line 83)
            Variable 'sigma_1' has been deleted from base workspace.
            Suggested Actions:
                • Undo the deletion. - Fix
                • Load a file into base workspace. - Fix
                • Create a new variable. - Fix
    Failure in initial objective function evaluation. LSQNONLIN cannot continue.

I don't understand why the two parameters results undefined since this method of using a function to calculate the error and give the function as input to the lsqnonlin, works in other context without a simulink model.

Do you know how to solve this problem? Is it right that the errorrs are caused by sigma_1 and Km_1? Maybe with a simulink model this is not the right way to use lsqnonlin?

Thank you!

Answer 1

[Torsten]

Did you study this example:

https://de.mathworks.com/help/optim/ug/lsqnonlin-with-a-simulink-model.html

?

[Raffaele]

Mh yes, but the simulink model I have uses the block "Translational Hard Stop" which has "sigma_1" and "Km_1" as input and I don't think this block allows input from input port as done in the example.

[Torsten]

I have no experience with Simulink, but input of parameters seems to work via

in = in.setVariable('Kp',pid(1),'Workspace',mdl);
in = in.setVariable('Ki',pid(2),'Workspace',mdl);
in = in.setVariable('Kd',pid(3),'Workspace',mdl);

doesn't it ?

So it seems natural to set

in = in.setVariable('sigma_1',p(1),'Workspace',mdl);
in = in.setVariable('Km_1',p(2),'Workspace',mdl);

in your case.

[Raffaele]

Yes, sorry this is my first time using simulink. It actually worked. Thank you!