Homotopy-based initialization does not use updated parameter values when re-simulating

[BuettR51]

It is not currently possible to obtain results for updated parameters in compiled models that require homotopy-based initialization. Instead, homotopy-based initialization always uses the default parameter values that the model was compiled with.

Please consider the following test case of a modified van der Pol oscillator:

model TestInitialization "Simple model to demonstrate homotopy bug"
   parameter Real A = 0.8;
   Real x;
   Real xPrime;
   Real mu;

initial equation
   x = 0.2;
   der(x) = -1;
equation
   der(x) = xPrime;
   mu = homotopy(0.1*x*der(x) + 0.1*der(xPrime), 0.1*der(x));
   der(xPrime) + mu*A*(x*x - 1)*xPrime + x = 0;
end TestInitialization;

Steps to reproduce the problem:

1. Simulate TestInitialization using –homotopyOnFirstTry.
2. Set the parameter A to 0.1 by using the Variables Browser, modifying Model_init.xml, using –override, or using –overrideFile.
3. Re-simulate the compiled model with -homotopyOnFirstTry using any method, then examine the results. The value of A will not be updated and the results will be the exact same as if they were run using A = 0.8. (If using the Variables Browser to re-simulate, the displayed value of A will be 0.1, but plotting it will show that it is actually 0.8 instead).

Performing the same steps without -homotopyOnFirstTry results in the parameter being updated as expected.

This behavior is present in more complicated models as well. It practically prohibits obtaining useful results from large-scale parameter sweeps where conventional initialization is likely to fail.

Thank you.

[casella]

@AnHeuermann, could you please have a look? This should be solved for 1.14.0 if possible.

[AnHeuermann]

I don't think I can solve this until Monday, since my priorities are on a different project at the moment.
I will have a look into it at the end of next week.

[casella]

OK, no problem

[casella]

Releasing 1.14.0 which is stable and has many improvements w.r.t. 1.13.2.

This issue, previously marked as blocker for 1.14.0, is rescheduled to 1.15.0

[casella]

Release 1.15.0 was scrapped, because replaceable support eventually turned out to be more easily implemented in 1.16.0. Hence, all 1.15.0 tickets are rescheduled to 1.16.0

[casella]

@AnHeuermann, can we fix this for 1.16.0?

[Karim.Abdelhak]

I changed something for homotopy resimulate in #6001, this could already be fixed. I will test tomorrow.

[AnHeuermann]

I'm busy with FMI at the moment. But maybe it is already fixed by @kabdelhak

[Karim.Abdelhak]

I tested with -override and it works perfectly fine. Adeel tested the other stuff for #6001, everything is fixed!

[casella]

Thank you folks!