how to keep gamultiobj goals positive

Hello,

I am trying to optimise a function using the GAMULTIOBJ tool, the function takes a 12x1 input vector and returns a 2x1 output vector, the input vector is the variables of the optimisation and the output are the goals that its trying to minimise. The problem is any time I try to run the optimisation the goals are negative numbers, I need to ensure that they stay positive. How do I put constraints on the goals alone but leave the input vector unconstrained?

Many thanks

Answer 1

[Alan Weiss]

If you have R2014b or later, use a nonlinear constraint function. Use one nonlinear constraint for each objective function. To save time, after debugging your constraints, you might want to use the technique in Objective and Nonlinear Constraints in the Same Function.

Alan Weiss

MATLAB mathematical toolbox documentation

[Steven Taggart]

Thanks Alan,

I will look into that. :)

[Steven Taggart]

Alan,

I was using a nonlinear constraint function, in order to keep x1 and x2 positive I created a function as follows

function [c,ceq] = mycon(x)  
c(1) = -x(1);
c(2) = -x(2);  
ceq = [];  
end

however it doesn't seem to be having an effect.

[Alan Weiss]

Don't use nonlinear constraints to constrain your decision variables to be positive. Use bounds for that. Use nonlinear constraints to bound nonlinear functions of the decision variables. See Types of Constraints.

Alan Weiss

MATLAB mathematical toolbox documentation

[Steven Taggart]

Hi Alan

Thanks for your help, does the bound constraint work on the output of the function? I thought they only bound the input variables vector.

Steven

[Alan Weiss]

No, bounds work on the decision variables, the x variables, not the output. I made the comment because you set

c(1) = -x(1)

which means your nonlinear constraint was trying to keep the input variable x(1) positive.

Alan Weiss

MATLAB mathematical toolbox documentation

[Steven Taggart]

Ah right, I using the GAMULTIOBJ tool, can you constrain the output to be positive?

I have achieved what I need with the FGOALATTAIN multi-objective tool but id rather use the GA one if possible.

Steven

[Alan Weiss]

As I stated in my original answer, if you have R2014b or later, then yes, gamultiobj accepts nonlinear constraints. Otherwise, no.

Alan Weiss

MATLAB mathematical toolbox documentation