Fmincon objective and nlcon shared simulink parameters

I'm trying to minimize a function of my vector X which is the input of my simulink model.

Right now i'm doing:

X=fmincon(@(X)objective,x0,A,B,Aeq,Beq,@(X)nonlcon,options)

function [f,gradf]=objective(X,simulation_param,Tfinal)
T=(0:1:Tfinal)';
TIN=[T X]
[~,~,yout]=sim('sim_model',[0 Tfinal],opt ,TIN);
f=some_fun_of(yout)
gradf=some_other_fun_of(yout)
end

function [G,P,gradG,gradP]=nonlcon(X,simulation_param,Tfinal)
T=(0:1:Tfinal)';
TIN=[T X]
[~,~,yout]=sim('sim_model',[0 Tfinal],opt ,TIN);
G=constr_fun_of(yout)
gradG=some_other_constr_fun_of(yout)
P=...
gradP=...
end

Since the simulation is the same, i would like fmincon to call it only once and not twice (once for objective fun and once for nonlcon fun).

I saw this link: https://www.mathworks.com/help/optim/ug/objective-and-nonlinear-constraints-in-the-same-function.html

but i dont know how to readapt the code if i have a Simulink simulation.

Answer 1

[Matt J]

function [x,f,eflag,outpt] = runobjconstr(x0,opts)

if nargin == 1 % No options supplied
    opts = [];
end

simulation_param=...,
Tfinal=...,

 xLast = []; % Last place update_yout was called
 yout=[];

fun = @objfun; % The objective function, nested below
cfun = @constr; % The constraint function, nested below

% Call fmincon
[x,f,eflag,outpt] = fmincon(fun,x0,[],[],[],[],[],[],cfun,opts);

    function [f,gradf] = objfun(x)
        if ~isequal(x,xLast) % Check if computation is necessary
            update_yout(x);
            xLast = x;
        end
        % Now compute objective function
        f=some_fun_of(yout)
        gradf=some_other_fun_of(yout)
    end

    function [G,P,gradG,gradP] = constr(x)
        if ~isequal(x,xLast) % Check if computation is necessary
            update_yout(x);
            xLast = x;
        end
     G=constr_fun_of(yout)
     gradG=some_other_constr_fun_of(yout)
     P=...,
     gradP=...,
    end
    
    function update_yout(X)
        
        T=(0:1:Tfinal)';
        TIN=[T X]
        [~,~,yout]=sim('sim_model',[0 Tfinal],opt ,TIN);        
        
    end

end

[Alberto Belvedere]

Thank you for your help!

If i want to provide also the hessian, that is a function of yout, to fmincon, can i proceed in the same way?

[Matt J]

Yes, you can do anything with yout

[Alberto Belvedere]

Since the hessian get passed through options i just make it point to a function that is nested? Like:

function runobjconstr(...)

options=optimoptions('fmincon',... ,'HessFcn', @nested_fun)

.

.

.

function hess=nested_fun(x,lambda)

end

end

[Matt J]

Yes, all functions nested within runobjconstr will have shared access to yout.

[Alberto Belvedere]

Thanks, i mark this as closed

[Farshid R]

Hello,

Could I ask a question about optimization?