simBiology exported model simulation output not in the same order as `InitialValues`

Hello,

I'm using a SimBiology exported model and I need to get the last value of a simulation and use it as initial condition for a following simulation (with different parameters, or dosing information), here is an example of what I need to do:

[t, y, names] = simulate(exported_model);
exported_model.InitialValues = y;
[t, y, names] = simulate(exported_model);

The problem is in the second line of the code as, surprinsingly, the order of the variables in y is not the same as the order in InitialValues. Currently, I'm performing a search for the appropriate index in names and using `exported_model.getIndex` to update each element of InitialValues. But this approach is suboptimal, it takes more than 1min to go over the entire array of initial conditions.

Is there any way to ensure that the output of model simulations come in the correct order? Or what would be the optimal approach here?

Thank you

[Valentina Vasic]

Hallo Paulo,

I have never tried it, I usually use your method, but I never needed to do double simulation. Maybe this can give you a hand. You have to go to the Examples.

An easier and faster method I don't think there is, but here is some advice.

Best,

Valentina

Answer 1

[Fulden Buyukozturk]

Hello,

exported_model.InitialValues lists all of exported_model's editable objects which is by default everything (compartments, species, parameters) whereas simulate(exported_model) returns states in the order of model's statesToLog, which is by default only non-constant quantities.

If you specify editable objects and statesToLog to be the same during model export, the order of quantities should be the same. Please see below for an example.

m = sbmlimport('lotka');
% define states for editable objects and statesToLog
states = sbioselect(m,'Name',{'y1','y2'});
% set statesToLog
m.getconfigset.RuntimeOptions.StatesToLog = states;
% export model specifying editable objects
em = export(m,states);
[t,y,names] = simulate(em);

{em.ValueInfo.Name}'

ans = 2×1 cell array
    {'y1'}
    {'y2'}

    
names

names = 2×1 cell array
    {'y1'}
    {'y2'}

Fulden

[Paulo]

Thank you for your answer,

Adding the following line did the trick:

m.getconfigset.RuntimeOptions.StatesToLog = states;

Answer 2

[Fangjun Jiang]

For a Simulink simulation, there are Inputs, Outputs and States. Usually, only States need initial values.

From "exported_model.InitialValues = y", it seems that you are assigning the Outputs of last simulation to the initial States of the next simulation. That is the problem. In general, Outputs will not be the same as the States, although they could be exactly the same.

There are ways to save the values of States and utilize it for the next simulation. See "States" and "Final states" in this doc.

web(fullfile(docroot, 'simulink/gui/data-import-export-pane.html'))

web(fullfile(docroot, 'simulink/ug/state-information.html'))

web(fullfile(docroot, 'simulink/ug/saving-and-restoring-simulation-operating-point.html'))

[Paulo]

Thank you, Fangjun, for you answer.

However, the solution you proposed is for Simulink and not SimBiology.

And yes, what I want to do is to use the Outputs of last simulation as initial state of the next, the two objects (y and InitialValues) do share references to the same state variables of the model, the only aspect that prevents line 2 to be valid, is the fact that, for some reason, SimBiology is not consistent in the order it stores the variables in these two arrays.

[Fangjun Jiang]

Based on the doc link below, it seems "names" in your code is the column labels of "y". I wonder if there is a column label for model.InitialValues.

If there is and the labels match but in a different order, then you could use ismember() to shuffle them to match.

https://www.mathworks.com/help/simbio/ref/simbiology.export.model.simulate.html

[Paulo]

Yes, that is my current approach (something similar to it anyway), I'm getting the index from names using:

    [~, i] = max(cellfun(@(x) strcmp(x, name), all_names));  

I do something similar using exported_model.getIndex, and I'm able to correctly update the initial values, but it takes more than one minute to do so. And that is even after I defined an array that maps one set of indexes into the other, so I don't have to the search every time.

So I was wondering if there is a more efficient way of doing it.

Thank you,