An adaptive genetic hyper-heuristic algorithm for a two-echelon vehicle routing problem with dual-customer satisfaction in community group-buying

This study focuses on a novel variant of the classical two-echelon vehicle routing problem (2E-VRP), termed the two-echelon vehicle routing problem with dual-customer satisfaction (2E-VRP-DS) (i.e. time windows satisfaction and freshness satisfaction) in community group-buying. It is important to obtain better solutions for the 2E-VRP-DS with long-distance distribution in the first echelon and last-mile delivery in the second echelon. Therefore, a new mathematical model is established for the 2E-VRP-DS that incorporates objectives: minimising the total distribution costs, and maximum dual-customer satisfaction (time windows satisfaction, and product freshness satisfaction). To solve the mathematical model, an efficient adaptive genetic hyper-heuristic algorithm (AGA-HH) was proposed, complemented by a k-means clustering approach to generate initial solutions. The adaptive genetic algorithm is considered to be a high-level heuristic, and ten local search operators were considered as low-level heuristics to expand the search region of the solution and achieve robust optimal results. Three sets of experiments were conducted, and the results demonstrated the superiority of AGA-HH in solving the 2E-VRP-DS, showing improvements in distribution costs reduction, time windows compliance, and product freshness preservation. Moreover, sensitivity analyses were carried out to show the influence of the number of DCs and the tolerance range of product freshness, discovering some managerial insights for companies. Future work should consider and investigate VRPs in other new business modes.