An estimation of distribution algorithm for the 3D bin packing problem with various bin sizes

Abstract

The 3D bin packing problem (3DBPP) is a practical problem modeled from modern industry application such as container ship loading and plane cargo management. Unlike traditional bin packing problem where all bins are of the same size, this paper investigates a more general type of 3DBPP with bins of various sizes. We proposed a modified univariate marginal distribution algorithm (UMDA) for solving the problem. A packing strategy derived from a deepest bottom left packing method was employed. The modified UMDA was experimentally compared with CPLEX and a genetic algorithm (GA) approach. The experimental study showed that the proposed algorithm performed better than GA and CPLEX for large-scale instances. © 2013 Springer-Verlag.