Multiple Non linear constraints for fmincon optimization

I would like to be able to use multiple non linear constraints with the fmincon optimization function. Currently fmincon works with both of my nonlcon function handles. Both of them are only using c(x) < 0, and not ceq(x) = 0.

So I have a function

[c,ceq] = cost(x, ..., ..., ...)

and corresponding function handle

nonlincon1 = @(x) cost(x, ..., ..., ...);

and

[c,ceq] = force(x, ..., ..., ...)

and corresponding function handle

nonlincon1 = @(x) force(x, ..., ..., ...);

Like I said I can pass the non linear constraint nonlincon1, or nonlincon2 with fmincon...

[xOpt,optimCost] = fmincon(f,x0,[],[],[],[],lb,ub,nonlcon1,options);

[xOpt,optimCost] = fmincon(f,x0,[],[],[],[],lb,ub,nonlcon2,options);

but I am not sure how to use both function handles as constraints at once. I have tried putting them in cells and the function works but I get some a message I do not quite understand, and the optimization is ignoring the nonlincon2 constraint.

[xOpt,optimCost,exitflag] = fmincon(f,x0,[],[],[],[],lb,ub,{nonlcon1;nonlcon2},options);

Warning: Constraint gradient function provided but OPTIONS.GradConstr = 'off'; ignoring constraint gradient function and using finite-differencing. Rerun with
OPTIONS.GradConstr = 'on' to use constraint gradient function.

Any ideas?

[Ronen Yudzinsky]

Yeah, this should be pretty easy.

function [c,ceq] = supernonlcon(x,..,n)
[c1 ceq1] = cost(x,..,n);
[c2 ceq2] = force(x,..,n);
c = [c1; c2];
ceq = [ceq1; ceq2]
end

[Joshua Hoegerman]

This doesnt work, you cant store multiple function handles within a standard array, you need to use a cell array which fmincon doesnt accept

[Matt J]

It should work, provided cost() and force() are visible to the function somehow, e.g., they are local functions.

Answer 1

[Alan Weiss]

The documentation addresses your problem: http://www.mathworks.com/help/toolbox/optim/ug/brhkghv-11.html#brhkghv-16

Alan Weiss

MATLAB mathematical toolbox documentation

Answer 2

[Jared]

Okay, so from the Warning above I have turned on the GradConstr option and I get an error,

 options = optimset('Algorithm','sqp','Display','iter','GradConstr','on');

 Error using fmincon (line 759)
 Gradient of nonlinear inequality constraints must have size 2-by-1.\

I think that I am heading down the wrong track with this....