Error with using a Simulink model as the objective function in ParetoSearch with options.UseParallel set to True (Global Optimization Tool Box)

Hi, I currently have an optimization problem which involves identifying a set of parameters in a Simulink model that finds the minimum of the model output. The output is the average of a time series data. I an currently using pareto search (it is a multi objective optimization problem) from the Global Optimization Toolbox.

I have a function handle (run_simulink_model()) that calls the simulink model and once the model runs based on the inputs (parameters of the model) given by the pareto search algo and then there is a bit of post processing that involves using the output from the model to compute the target for optimization. Now the whole thing runs fine when the options.UseParallel is set to . When I set options.UseParallel to the pareto search fails. In the case when parallel is set to I use parsim inside the function handle and it still doesnt seem to work. I would like to understand if parallel computing is possible in pareto search when a simulink model is used.

function fval = run_simulink_model(model_params_fr_optimization, nameSimulinkModel, save_location, evaluation, parallel)

batchSize = size(model_params_fr_optimization, 1);

fval = nan(batchSize, 3);

save_system(nameSimulinkModel);

simulationModelIn(batchSize) = Simulink.SimulationInput;

for idx1 = 1:batchSize

simulationModelIn(idx1).ModelName = nameSimulinkModel;

simulationModelIn(idx1) = simulationModelIn(idx1).setVariable(...

'param_name', model_params_fr_optimization(idx1,:));

end

if parallel == false

youts = sim(simulationModelIn);

elseif parallel == true

youts = parsim(simulationModelIn, 'ShowProgress', 'on', 'TransferBaseWorkspaceVariables','on');

end

fval = post_processing(youts) % Post processing in this section to return the multiple outputs to be optimized

end

%% Function handle and the script for paretosearch

f = @(x)run_simulink_model(x, nameSimulinkModel, save_location, evaluation, parallel_run);

ptions = optimoptions('paretosearch');

options.Display = 'iter';

options.MaxFunctionEvaluations = 10;

options.MaxIterations = 5;

options.InitialPoints.X0 = X; %initial data points

options.ParetoSetSize = 4;

options.UseParallel = true;

[X,fval,exitflag,output,residuals] = paretosearch(f, 2, [], [], [], [], lb, ub, [], options)

Following is the error message i get:

Invalid Simulink object name: simpleSimulinkModel.

The block diagram 'simpleSimulinkModel' is not loaded.

If you need nay further info please do let me know. :) TIA

Answer 1

[Avadhoot]

Hi Jyotirmaya,

This issue appears because MATLAB cannot find your model in the base workspace before simulating it. You can solve this problem using any of the following steps:

1. Load the model: Use the following line to explicitly load the model file in the base workspace:

load_system(nameSimulinkModel);

Insert this line instead of the following line

save_system(nameSimulinkModel);

2. Check if the model is on the MATLAB path: Execute the following command to add the model to the MATLAB path

addpath('path_to_the_model');

3. If both of the options don't work, ensure the file is open by excuting the following command:

open_system(nameSimulinkModel);

If the file opens without any issuesd then try re-running your script.

I hope it helps.