Multi-drop container loading using a multi-objective evolutionary algorithm

We describe a new algorithm MOCL (multiobjective container loading) for the multi-drop single container loading problem. MOCL extends the recent biased random-key genetic algorithm due to Goncalves & Resende to the multidrop problem by enhancing its genetic representation, its fitness calculations, and its initialisation procedure. MOCL optimises packings both for volume utilisation and for the accessibility of the packed objects, by minimising the number of objects that block each other relative to a pre-defined unpacking schedule. MOCL derives solutions that are competitive with state-of-the-art algorithms for the single-drop case (where blocking is irrelevant), plus it derives solutions for 2-50 drops that give very good utilisation with no or very little blocking. This flexibility makes MOCL a useful tool for a variety of 3D packing applications.