how to use FAST RESTART for different simulink model instances?

Hi together,

I'm having a problem with fast restart. The problem is stated as following: I have a Simulink model for a parameter sweep study. For each parameter value, the model will be tested under six different scenarios. To avoid repetitive compiling process, I would like to utilize fast restart to speed up the iteration of parameter sweep. The pseudo code can look like this:

simIN(1:6) = Simulink.SimulationInput("my_model"); % generate six model instances, corresponding to six test scenarios

for id = 1:6
    simIn(id) = simIn(id).setVariable(trajectory_id); % modify each instance to simulate test scenario 1-6
end

for it = 1:100
    simIn(1:6) = simIn(1:6).setVariable(parameter_it); %parameter sweep
    out = parsim(simIN,'UseFastRestart','on'); % parallel simulation for six scenarios
    cost = J(out); % calculate cost function
    parameter(it+1) = decide_next_parameter_to_evaluate;
end

min(J(out)) % get the optimum parameter with min cost function

I have noticed that, with the FASTRESTART, actually the process of the parallel simulation of scenario 1-6 was sped up. But during the parameter sweep (for it = 1:100), the model was constantly recompiled and terminated. What I want to achieve is to enable FASTRESTART for each model instance during its parameter sweep. To put it more intuitively, each model instance is compiled once and get an "individual" FASTRESTART during the parameter sweep. Do you have any recommendations? Thanks a lot in advance!

(In addition, the parameters of the sweep study cannot be generated in advance (due to bayesian dependency), so that I could not directly do parallel simulation to the 100 parameters. That means, the six scenarios have to be tested as a bundle, and after that, continue with the next parameter.)

Answer 1

[Ayush]

Hey,

To efficiently use "Fast Restart" in your Simulink model for parameter sweeps across different scenarios without recompiling the model every time, you can follow the following steps:

1. Pre-load and set up your model for Fast Restart.
2. Adjusting the loop structure for parameter sweeos and scenario simulations.

Here is the revised pseudo code for the above approach:

load_system('my_model'); % Load the model into memory
set_param('my_model', 'FastRestart', 'on'); % Enable Fast Restart
for it = 1:100
    parameter_it = decide_next_parameter_to_evaluate(); % Determine the next parameter value

    for id = 1:6
        simIn(id) = Simulink.SimulationInput('my_model');
        simIn(id) = simIn(id).setVariable('trajectory_id', id); % Set scenario-specific variable
        simIn(id) = simIn(id).setVariable('parameter_it', parameter_it); % Set parameter for current iteration
    end

    out = parsim(simIn, 'UseFastRestart', 'on'); % Run simulations in parallel with Fast Restart
    cost = J(out); % Calculate cost function
end

% Clean up by turning off Fast Restart
set_param('my_model', 'FastRestart', 'off');

For more information on "Fast Restart" in Simulink, refer to the following MathWorks documentation link: https://in.mathworks.com/help/simulink/run-individual-simulations.html

Hope this helps!

[stdrj]

Hi Ayush,

thanks for the answer. But this would not solve my problem, because parsim must be reinitialized for each iteration and "forgets" what the model is. Actually I found another solution, which is to

1. generate six hard copies of a model
2. give them the proper parameter for the scenario simulation
3. and enable fast restart for each of them

During the iteration, I just simulate with six copies of models with the efficiency guaranteed by Fast Restart. This is fast enough for me.