A forward-looking matheuristic approach for the multi-period two-dimensional non-guillotine cutting stock problem with usable leftovers

Abstract

In Birgin et al. (2020) the multi-period two-dimensional non-guillotine cutting stock problem with usable leftovers was introduced. At each decision instant, the problem consists in determining a cutting pattern for a set of ordered items using a set of objects that can be purchased or can be leftovers of previous periods; the goal being the minimization of the overall cost of the objects up to the considered time horizon. Among solutions with minimum cost, a solution that maximizes the value of the leftovers at the end of the considered horizon is sought. A forward-looking matheuristic approach that applies to this problem is introduced in the present work. At each decision instant, the objects and the cutting pattern that will be used is determined, taking into account the impact of this decision in future states of the system. More specifically, for each potentially used object, an attempt is made to estimate the utilization rate of its leftovers and thereby determine whether the object should be used or not. The introduced approach is compared with an exact off-the-shelf commercial solver and a myopic technique. Numerical experiments show the efficacy of the proposed approach.