Write nonlinear constraint to fmincon

I am using fmincon to optimize a vectorized function (myfun) of five variables (x). This function has also some parameters (k, l, m, n) which I am passing to fmincon. What I am doing is shown below:

p=@(x)myfun(x, k, l, m, n);
[xmin, pmin]=fmincon(p, x0, A, b, Aeq, beq, lb, ub, nonlcon, options);

My question is how am I going to write the nonlinear equality which I have in my problem. I have written another function about the nonlinear equality in which x is the variables used in myfun and L is a parameter that I would like to pass inside the constraint. See below:

function [c,ceq]=pitch(x,L)
c=[];
ceq=x(3)*x(5)-L;
end

The problem is that I don't know how to write inside the fmincon the constraint. I have written this:

nonlcon=@(x)pitch(x,L)

which results in an error. I would appreciate your answers. Thank you.

[Matt Kindig]

Could you post the error for us?

Answer 1

[Shashank Prasanna]

Thanks for the error, it appears that your objective function myfun takes 'x' as a 4x1 vector while your nonlinear constraint assumes that there is atleast x(5) in the line:

ceq=x(3)*x(5)-L;

that is atleast a 5x1 vector.

You might want to check that.

Answer 2

[Giorgos Papakonstantinou]

Attempted to access x(5); index out of bounds because numel(x)=4.

Error in power_fraction2 (line 7) Pf=(2*Rf*jmpp*x(2)^3*(x(3) - x(1))^2)/(3*x(3)*x(5)*x(3)*Vmpp*x(1));

Error in @(x)power_fraction2(x,jmpp,Vmpp,Rito,Rb,Rf,rc)

Error in fmincon (line 601) initVals.f = feval(funfcn{3},X,varargin{:});

Error in opt_power (line 48) [xmin, pmin]=fmincon(p, x0, A, b, Aeq, beq, lb, ub, nonlcon, options);

Caused by: Failure in initial user-supplied objective function evaluation. FMINCON cannot continue.

[Giorgos Papakonstantinou]

That's the error. Thank you.

Answer 3

[Giorgos Papakonstantinou]

Thank you Benji for the answer. This would make sense if I had not defined x(5) in the function that I would like to optimize. I am posting an equation included in myfun as an example.

Pf=(2*Rf*jmpp*x(2)^3*(x(3) - x(1))^2)/(3*x(3)*x(5)*x(3)*Vmpp*x(1));

So it is clearly stated in the function that vector x has 5 elements. What do you think?

Answer 4

[Giorgos Papakonstantinou]

I am sorry. Indeed I had a mistake in the boundaries. I had four elements instead of 5.

Answer 5

[Giorgos Papakonstantinou]

I would like one of the optimized variables to be an integer. Is there a way to achieve this with fmincon. In global optimization toolbox the Genetic Algorithm Solver can achieve that as far as I know. Am I correct or this can be done through fmincon?

[Matt J]

No, fmincon cannot do discrete optimization.

[Giorgos Papakonstantinou]

So Matt I have to use the Genetic Algorithm.

[Shashank Prasanna]

Giorgos, you are right this options is currently only available in Genetic Algorithm. Other optimization tools as yet dont support mixed integer programming.