Rule-based design of a granular tabu search for the multi-depot vehicle routing problem

We describe a rule-based process to design a granular tabu search for the multi-depot vehicle routing problem. Using this process, only known and simple algorithmic components from the literature are required to achieve a reasonable performance compared to state-of-the-art metaheuristics. In each step, we analyze a single algorithmic component by comparing different settings for this component. The selection of the best setting for each component relies on two clear rules that are based on the Wilcoxon signed-rank test and a straightforward tradeoff measure. By either focussing on quality or on speed, we derive two final granular tabu search configurations. The decisions on the algorithmic components are partly in line with previous recommendations for other types of vehicle routing problems from the literature but also differ in some important aspects.