fuzzy tuning error using a custom cost function

after runing this command to tune

fisRuleTuned = tunefis(FIS1,rule,fcn,options);

i am getting these errors

Error using costFcn

Unrecognized method, property, or field 'signals' for class

'Simulink.SimulationData.Dataset'.

Error in @(fis)costFcn(fis,model,minLevel,refLevel,wsVars,minData)

Error in fuzzy.tuning.internal.fisFitnessFcnWithoutIntCon>localFcn (line 54)

fitness = fitnessFcn(fis);

Error in fuzzy.tuning.internal.fisFitnessFcnWithoutIntCon (line 25)

fitness(i) = localFcn(params(i,:),fis,spec,fitnessFcn,discSet, ...

Error in fuzzy.tuning.internal.optimFISWithGA>@(x)fuzzy.tuning.internal.fisFitnessFcnWithoutIntCon(x,pdata.fis,pdata.spec,evalFcn,pdata.discset,options.IgnoreInvalidParameters,false,options.TuningOutputSinks) (line 70)

p.fitnessfcn = @(x)fuzzy.tuning.internal.fisFitnessFcnWithoutIntCon(...

Error in createAnonymousFcn>@(x)fcn(x,FcnArgs{:}) (line 11)

fcn_handle = @(x) fcn(x,FcnArgs{:});

Error in makeState (line 58)

firstMemberScore = FitnessFcn(state.Population(initScoreProvided+1,:));

Error in galincon (line 24)

state = makeState(GenomeLength,FitnessFcn,Iterate,output.problemtype,options);

Error in ga (line 420)

[x,fval,exitFlag,output,population,scores] = galincon(FitnessFcn,nvars, ...

Error in fuzzy.tuning.internal.optimFISWithGA (line 89)

[params,fval,exitflag,output,population,scores] = ga(p);

Error in tunefis (line 326)

[varargout{1:nargout}] = fuzzy.tuning.internal.optimFISWithGA(pdata,fitnessFcn,options,kFoldData);

Caused by:

Failure in initial user-supplied fitness function

evaluation. GA cannot continue.

the cost function i have used below

function cost = costFcn(fis,model,minLevel,refLevel,wsVarNames,varargin)

assignin('base',wsVarNames,fis)

out = sim(model);

glucose = out.yout.signals(1).values;

insulin = out.yout.signals(2).values;

tout = out.yout.time;

err = glucose - refLevel;

err(glucose<minLevel) = 100;

errSquare = err.^2;

meanSquare = mean(errSquare);

cost = sqrt(meanSquare);

if isempty(varargin)

return

end

data = varargin{1};

if cost < data.MinCost

data.MinCost = cost;

data.MinGlucose = glucose;

data.MinInsulin = insulin;

data.fisMin = fis;

data.Tout = tout;

end

end

[Sam Chak]

@N Naveen

I'm unsure. Can you really pass the Simulink model when creating the Custom Cost Function?

Have you tried extracting the data from out.yout.signals, and then pass them to the CostFcn as trainingData?

function cost = costFcn(fis, model, minLevel, refLevel, wsVarNames, varargin)

    assignin('base', wsVarNames, fis)
    out        = sim(model);
    glucose    = out.yout.signals(1).values;
    insulin    = out.yout.signals(2).values;
    tout       = out.yout.time;
    err        = glucose - refLevel;
    err(glucose<minLevel) = 100;
    errSquare  = err.^2;
    meanSquare = mean(errSquare);
    cost       = sqrt(meanSquare);
    
    if isempty(varargin)    
        return
    end
    
    data = varargin{1};
    
    if cost < data.MinCost
        data.MinCost    = cost;
        data.MinGlucose = glucose;
        data.MinInsulin = insulin;
        data.fisMin     = fis;
        data.Tout       = tout;
    end
    
end

Answer 1

[Akshat Dalal]

Hello Naveen,

I understand you are facing some errors in using a custom cost function while tuning a fuzzy system using the Fuzzy Logic Toolbox.

I believe your custom cost function is inspired from the Fuzzy Logic Toolbox example provided in documentation - https://in.mathworks.com/help/releases/R2023a/fuzzy/design-fuzzy-logic-controller-for-artificial-pancreas.html

Assuming that is the case, I ran this example using your custom cost function and it did not produce any errors. I believe this error is occurring due to a discrepancy in your model. You could refer the example documentation for more information.