Genetic Algorithm Custom Output

Hello, I'm trying to optimize a simple Simulink model output by using Genetic Algorithm. Three different output values are implemented in this model. The first one is used as cost value. Hovewer, the other two values are not assigned to anything. I want to get these outputs as timeseries after the Genetic Algorithm calculations, like "The second output of Generation XX Individual YY is ...". When the calculations are done with single core, I can get all the results I want, but when the parallel toolbox is used, only the last generation and last individual is visible after the calculations. How can I fix that parallel calculation? Here is my simple code that gives all individuals and cost values for me, but I need the other two outputs of simulation. Thanks.

clear all;clc;,
%%
tic
saveFlag = 0;
rng('shuffle');
objectiveFunction = @(params) nlaSimResults(params);

lowerBound = [1000 0.2];
upperBound = [25000 0.8];

popSize = 10;
genSize = 20;

options = optimoptions('ga','UseParallel',true,'UseVectorized',false,'PopulationSize',popSize,'MaxGenerations',genSize,'OutputFcn',@gaOutputFunction,'Display','iter');

[bestParams, minCost, ~, optResults, optPop, optScores] = ga(objectiveFunction,2,[],[],[],[],lowerBound,upperBound,[],options);
optimTime = toc;

if saveFlag == 1; save('OptimResulst','bestParams','minCost','optResults','optScores','optimTime','popSize','genSize'); end

function cost = nlaSimResults(params)
    global otherVars
    
    Altitude = params(1);
    Mach = params(2);

    assignin('base','Altitude',Altitude);
    assignin('base','Mach',Mach);

    sim('debugModel',10); % Simulation Model with Altitude and Mach inputs
    cost = logsout{1}.Values.Data(1); % The first output of the model is assigned as cost
    
    if isempty(otherVars)
        otherVars = {};
    end

    otherVars = [otherVars; logsout{2}.Values.Data logsout{3}.Values.Data];

    assignin('base','otherVars',otherVars);
end

function [state, options, flag] = gaOutputFunction(options, state, flag)
    persistent popHistory bestValues bestIndividuals otherVarsHistory
    global otherVars

    if isempty(popHistory)
        popHistory = {};
        bestValues = [];
        bestIndividuals = [];
        otherVarsHistory = {};
    end

    [bestScore, bestIdx] = min(state.Score);
    bestValues = [bestValues; bestScore];
    bestIndividuals = [bestIndividuals; state.Population(bestIdx, :)];

    popHistory(state.Generation + 1) = {state.Population};

    otherVarsHistory{state.Generation + 1} = otherVars;
    otherVars = [];

    assignin('base','popHistory',popHistory);
    assignin('base','bestValues',bestValues);
    assignin('base','bestIndividuals',bestIndividuals);
    assignin('base','otherVarsHistory',otherVarsHistory);
end

My model is so simple, just returning the sum, product and division of the altitude and mach values.

Answer 1

[Alan Weiss]

I am not sure, but perhaps using the workflow in this example will help:

Optimize Simulink Model in Parallel

Alan Weiss

MATLAB mathematical toolbox documentation