Quadprog solver not following constraints

Hello!

I am working on a project where I use quadprog to solve an optimization problem for a model predictive controller (MPC). I am trying to apply very simple bounds to the optimization problem but it does not seem to be working.

Here is a snippet of my code:

for i =1:2*iC
    lb(i,:)=-50000;
    ub(i,:)=50000;
end

duf = quadprog(H,F,[],[],[],[],lb,ub);

The for loop ensures that lb and ub are the same size as duf. However, when I run the program, quadprog seems to simply ignore the constraints. Am I missing something simple? Thank you!

[Nathan Batta]

It does follow bounds that are equal to zero though. So if either the upper or lower bound is zero, it will follow that constraint but nothing else

[Matt J]

We need a description of what you actually see. What indications do you see that the bounds are not obeyed? How much larger is the violation than the ConstraintTolerance parameter? Also, please attach a .mat file containing H, F, lb, and ub so that we can run the optimization ourselves and try to reproduce your output.

Answer 1

[Abdolkarim Mohammadi]

Your bounds are the same for all of the variables, so they can be defined easier:

lb = -5e4 * ones(1,2*iC);
ub =  5e4 * ones(1,2*iC);

[Nathan Batta]

Good catch! However, I am still having the same issue.

[Abdolkarim Mohammadi]

Have you tried increasing the MaxIterations option?

[Nathan Batta]

Yes. It didn't have any effect on the optimization