Adaptive constrained multi-objective differential evolution algorithm for vehicle routing problem considering crowdsourcing delivery☆

With the increase of logistics orders, the crowdsourcing delivery with part-time drivers is an effective way to solve urban logistics distribution. The crowdsourcing distribution is the vehicle routing problem containing service time windows of customers and delivery time windows of part-time drivers. However, it is challenging to obtain optimal scheduling schemes for the vehicle routing problem considering crowdsourcing (VRP-C). To address this constrained optimization problem, an adaptive constrained multi-objective differential evolution (ACMODE) algorithm is designed in this paper to minimize the travel distance and the driver payment. First, a two-stage initialization method is designed to generate solutions with fewer constraint violations by dual time windows operation. Second, a neighborhood-oriented search strategy is developed to guide searching more feasible regions and avoiding the premature convergence of solutions. Third, a fast selection mechanism based on an improved non-dominated sorting approach is proposed to achieve the tradeoff on objectives and constraints, accelerating the process of optimization. Finally, several numerical simulation experiments explain that the proposed ACMODE algorithm can obtain feasible solutions of the constrained multi-objective optimization problem effectively, and has better performance than some state-of-art algorithms in solving VRP-C.