Considerations on using genetic algorithms for the 2D bin packing problem: A general model and detected difficulties

Abstract

The 2-dimensional bin packing problem appears in various fields across many industries such as wood, glass, or paper industries. They may differ in terms of specific constraints with respects to each area but they all share a common objective that is to maximize the material utilization. Belonging to the class of NP-Hard problems, there exist no efficient method to solve it, but only approximate solution by combining a greedy placement strategy with some optimization techniques such as Genetic Algorithms. That combination approach is very popular in this topic, but few researchers have clearly presented their method and no one has explained the difficulty of applying Genetic Algorithms to this problem, making it difficult for new researchers to reimplement the known algorithms. In this paper, in addition to proposing a general framework for applying Genetic Algorithms to solving this problem, we also identify the main difficulties of using this approach and propose 2 genetic operators, path recombination and hill-climbing mutation, to support our genetic model.