ABSTRACT. This paper studies a variant of the container loading problem in which to the classical geometric constraints of packing problems we add other conditions appearing in practical problems, the multidrop constraints. When adding multi-drop constraints, we demand that the relevant boxes must be available, without rearranging others, when each drop-off point is reached. We present first a review of the different types of multi-drop constraints that appear in literature. Then we propose a GRASP algorithm that solves the different types of multi-drop constraints and also includes other types of realistic constraints such as full support of the boxes and load bearing strength. The computational results validate the proposed algorithm, which outperforms the existing procedures dealing with multi-drop conditions and is also able to obtain good results for more standard versions of the container loading problem.

No abstract

No abstract

This paper proposes a matheuristic algorithm based on a column generation structure for the capacitated vehicle routing problem with three-dimensional loading constraints (3L–CVRP). In the column generation approach, the master problem is responsible for managing the selection of best-set routes. In contrast, the slave problem is responsible for solving a shorter restricted route problem (CSP, Constrained Shortest Path) for generating columns (feasible routes). The CSP is not necessarily solved to optimality. In addition, a greedy randomized adaptive search procedure (GRASP) algorithm is used to verify the packing constraints. The master problem begins with a set of feasible routes obtained through a multi-start randomized constructive algorithm (MSRCA) heuristic for the multi-container loading problem (3D–BPP, three-dimensional bin packing problem). The MSRCA consists of finding valid routes considering the customers' best packing (packing first-route second). The efficiency of the proposed approach has been validated by a set of benchmark instances from the literature. The results show the efficiency of the proposed approach and conclude that the slave problem is too complex and computationally expensive to solve through a MIP.