The container transshipment problem: Searching representation landscapes with metaheuristic algorithms

The container transshipment problem involves scheduling a fleet of lorries to collect and deliver containers of various sizes while minimizing the total distance travelled. The problem originates in the need for logistics companies to solve the problem on a regular basis as part of their daily operations. In this paper, we compare two genetic algorithms tailored to solve this problem based on permutation and bin-packing inspired encodings. Results are presented and analysed in order to evaluate the validity and robustness of the two approaches. As part of the analysis, bounds were calculated to determine how well both GAs perform in absolute terms as well as relative to each other. Of the two GAs there is one clear winner, although it is not the one that would have been indicated by previous research. Whilst the winning GA is able to generate significant savings in practice, compared to the optimum there remains room for further improvement.