Is it possible to use 'trust-region' algorithm for the fsolve function inside a Simulink model?

I have a simulink model that contains a MATLAB function block. This matlab function block recalls the function fsolve, the optimization options chosen for the fsolve function are as follow:

options = optimoptions('fsolve','Display','none','Algorithm','levenberg-marquardt');
[S,fval,exit,output] = fsolve(f,S_0,options);

This works fine but i want to use for the algorithm value 'trust-region' instead of 'levenberg-marquardt', because it gives results that are more precise:

options = optimoptions('fsolve','Display','none','Algorithm','trust-region');

Trying to run my model in simulink with this line of code gives me this error:

OPTIMOPTIONS property 'Algorithm' must be one of the following values for fsolve:

'levenberg-marquardt'

Function 'resolve_S2.m' (#574.167.233), line 4, column 11: "optimoptions('fsolve','Display','none','Algorithm','trust-region')"

I should say that the MATLAB code works when not connected to the simulink environment with both options for the algorithm value, it only stops working when using the 'trust-region' option in simulink.

Is there a way to have the 'trust-region' option accepted by the simulink environment?

Answer 1

[Kartik]

Hi Marco,

As of R2022b, Simulink supports only 'levenberg-marquardt' algorithm for “fsolve”.

It is possible using coder.extrinsic, but then code generation is not possible.