Integer optimization returns decimal numbers

Hello!

I am working with a problem where I have implemented an integer linear programming optimization. The optimization variables are all integers with the following setup:

P = optimvar('P', 24,5,'Type','integer','LowerBound',0,'UpperBound',5);

alfa = optimvar('alfa' ,24,5,'Type','integer','LowerBound',0,'UpperBound',1);

I have added an objective function and a set of active constraints, and an optimal solution can be found in every iterations of my algorithm.

Although, in the results I can notice that the optimization variables sometimes are chosen as decimal numbers, e.g. 1e-5 or 1e-15.

How come this can happen?

BR

Johannes

Answer 1

[Alan Weiss]

The intlinprog solver (which solve calls internally for MILP problems) does not necessarily return or use internally exact integers. If you want exact integers, round your results, which might result in some linear constraints becoming slightly infeasible. See Limitations.

Alan Weiss

MATLAB mathematical toolbox documentation