Optimization problem on Constraint declaration

Hello,

I have a code

syms a wd;
c4_update=0.0248*abs(a)^4 - 9.878; %coming from some other equation
c3_update=abs(wd*(- 0.02+ a*(0.040 - 0.1352i)) + 0.223 - 0.223i)^2 + 0.223+ 0.223i)^2 + a*(0.096- 0.089i)) + 0.22 - 0.223i)^2 % coming from some other equation
nonlcon =@constraintfcn;
[x_value2,fval2]  = fmincon(fun,x0,A,b,Aeq,beq,lb,ub,nonlcon2,opts); % where  fun,x0,A,b,Aeq,beq,lb,ub,opts are well defined

function [c,ceq] = constraintfcn(c3_update,c4_update)
c(1)= c3_update(1);
c(2)=c4_update(1);
ceq = [];
end

It gives error. How to solve this problem.

[Matt J]

It gives error.

An error you haven't shown us...

[Soumili Sen]

It shows "FMINCON requires all values returned by functions to be of data type double."

[Matt J]

Yes, but you haven't copy/pasted the full error message, so we cannot see where the error is coming from.

Answer 1

[Matt J]

We can't tell from your code what the relation is supposed to be between a, wd and x. This might be what you were trying to do, however:

nonlcon =@constraintfcn;
[x_value2,fval2]  = fmincon(fun,x0,A,b,Aeq,beq,lb,ub,nonlcon2,opts); % where  fun,x0,A,b,Aeq,beq,lb,ub,opts are well defined

function [c,ceq] = constraintfcn(x)

  a=x(1); wd=x(2);
  
  c(1)= abs(wd*(- 0.02+ a*(0.040 - 0.1352i)) + 0.223 - 0.223i)^2 + 0.223+ 0.223i)^2 + a*(0.096- 0.089i)) + 0.22 - 0.223i)^2;
  c(2)=0.0248*abs(a)^4 - 9.878;
  
  ceq=[];
end

[Soumili Sen]

ok understand. But one problem is that the values of c4_update and c3_update are random in nature (which is coming from other calculation) which I want as the input of the function. I can't define c(1) and c(2) like that,coz it will change for next calculation.

[Matt J]

For any fixed value of x, your constraint function constraintfcn(x) must return the same value all the time. The defintiion of the function is not allowed to evolve, stochastically or otherwise, throughout the optimization.

[Soumili Sen]

ok. It's mean I can't use c4_update and c3_update as an input,right? I have to use a fixed value (like x)?

[Matt J]

Both the constraint and objective functions must be a function of your unknown variable vector x (and nothing else). You can make the definition of the constraint function depend on other fixed problem parameters, however that definition must remain fixed throughout the optimization. During the optimization, the only thing that is allowed to vary the output of the constraint and objective functions are the different guesses for the unknowns that fmincon will iteratively pass in as input.

[Soumili Sen]

Ok understand. Thank you for your valuable suggestions.