How to avoid redefining optimoptions in Simulink for real-time nonlinear optimization with fmincon (SQP)?

Hello guys, I'm trying to implement a real-time nonlinear optimization solver using fmincon with the SQP algorithm in Simulink. I'm currently using a Function (Fcn) block and wrote the below optimization code inside it:

function y = fcn(u)
options = optimoptions('fmincon','Display','None','Algorithm','sqp');
x0=0.1;
lb=-10;
ub=10;
Cost_Function = @(x) f1(x,u);
y = fmincon(Cost_Function,x0,[],[],[],[],lb,ub,[],options);
end

function cost = f1(x,u)
cost=abs(x^2-u^2);
end

The code works correctly, but it's too time-consuming for my application, as the optimization needs to be performed frequently. One of the performance bottlenecks seems to be the repeated definition of "optimoptions" inside the block.

To address this, I tried defining the "options" object just once using "persistent" and "global" variables. However, Simulink throws an error, likely because options is not recognized as a Simulink parameter.

I also experimented with a "MATLAB System block" instead of the Fcn block, but the same issue persists — I can't define "optimoptions" only once at the start.

Question:

Is there a way to define optimoptions once (e.g., during initialization) and reuse it in Simulink for real-time execution, without redefining it at every simulation step?

Any suggestions or best practices for improving the performance of fmincon-based optimization in Simulink would be appreciated.

[Matt J]

Is your optimization problem just a fake placeholder example? I don't see why you would be using fmincon for such a thing.The solution is obviously x=±u.

In any case, for a 1D problem, you should just use fminbnd which won't require optimoptions, if all you're trying to do is to suppress termination messages.

u=6;
lb=-10;
ub=10;

y = fminbnd(@(x) f1(x,u),lb,ub)

y = -6.0000

function cost = f1(x,u)
 cost=abs(x^2-u^2);
end

Answer 1

[Paul]

Are you really using a Matlab Function block?

If so have you tried:

function y = fcn(u)

persistent options
if isempty(options)
options = optimoptions('fmincon','Display','None','Algorithm','sqp');
end

x0=0.1;
lb=-10;
ub=10;
Cost_Function = @(x) f1(x,u);
y = fmincon(Cost_Function,x0,[],[],[],[],lb,ub,[],options);
end

function cost = f1(x,u)
cost=abs(x^2-u^2);
end

If that code in the Matlab Function block throws an error, please show the full text of the error message.

If not, please provide a link to the specific block called "Matlab (fcn)"

Answer 2

[Matt J]

You can use global variables (but with care).

function y = fcn(u)
 
 global lb ub options

 y = fminbnd(@(x) f1(x,u),lb,ub,options)

end

function cost = f1(x,u)
 cost=abs(x^2-u^2);
end