A framework algorithm for a real-world variant of the vehicle routing problem

The vehicle routing problem has a half-a-century development history with many variants and algorithms. In this work, we tackle a real-world variant which has a heterogeneous fleet of vehicles and requires scheduling at limited-slotted depots with a time window constraint at a centralized garage. We first give problem formulations, introduce our approach to decompose this variant into subproblems and present our framework algorithm with experiments on real-world data sets. Experimental results show that our algorithm outperforms the heuristic method relaxed from dynamic programming and the multi-stage local search method. Our result is also far better than that of the experience-based greedy method. Significant savings with low resource consumption suggest practical use of our algorithm in real-world applications.