Simulation with large Data and Simulation​Data.Datas​etRef Objects; Avoiding eval

Hi everyone,

for my PhD project I need to run simulations in Simulink. In a Monte Carlo Approach I want to simulate many inputs (order of several 100) that I randomly generate. As these Signal Timeseries' require a lot of Memory, so I opted for using DatasetRef Objects: https://de.mathworks.com/help/simulink/slref/simulink.simulationdata.datasetref-class.html

As also described here: https://de.mathworks.com/help/simulink/ug/stream-input-data-for-a-parallel-simulation.html I want to stream my Data. In the example however they have their Datasets for simulations named Road1, Road2 etc, which I have read is bad practice, so I want to avoid it.

I already learned, that I can save my data as a Cell Array of datasets that each contain a timeseries with input data. However I cannot create DatasetRef Object from those entries to hand over to parsim, as the Simulink.SimulationData.DatasetRef() function requires a variable name as an input, but does not accept "Data{i}" as a variable name.

Unfortunately the Data is to large for upload but my code would look like this:

Store = matfile(append(path,"test.mat"));
numTestCases=length(Store.Data);
for i=1:numTestCases    
    inputData(i)=Simulink.SimulationData.DatasetRef(append(path,"test.mat"),sprintf("Data{%i}",i));
end

And Store beeing:

Store = 

  matlab.io.MatFile

  Properties:
              Properties.Source: 'path\test.mat'
            Properties.Writable: false                                                                                                                  
    Properties.ProtectedLoading: false                                                                                                                  
                           Data: [20x1 cell]                                                                                                            

  Methods
  
  

Has someone here an idea how to avoid using eval to just create the datasets as Data1,Data2 etc?

Cheers and thank you

[Alvaro]

You can look at what is loaded in the example given in the documentation that you linked by executing

structExample = load('suspn_3dof_test_cases.mat')

and you can see that it is not a cell array of datasets but a struct with fields Road1, Road2, etc.

It is possible to dynamically add fields to structs:

https://www.mathworks.com/help/matlab/matlab_prog/generate-field-names-from-variables.html

For example, see the struct generated by this code.

for i = 1:10
    road = strcat('Road',num2str(i));
    myStruct.(road) = rand(1,3);
end 

Could you explain further why generating a struct in this manner would cause any issues?

Alternatively, you could consider converting your cell array to an struct using cell2struct.

https://www.mathworks.com/help/matlab/ref/cell2struct.html

Answer 1

[Smattering_Applause]

See the Alternative method within this link: https://www.mathworks.com/matlabcentral/answers/292111-how-can-i-save-the-contents-of-a-struct-to-a-mat-file#answer_226587

If you can make it a cell array you can make it a struct, then save the struct in that method and load and then you'll have variable names without eval.