Equity-oriented vehicle routing optimization for catering distribution services with timeliness requirements

Abstract

The rapid development in the catering services and urban logistics industries has significantly promoted the prosperity of business-to-consumer (B2C) e-commerce urban logistics distribution, which is gaining increasing interest from food producers, distribution platforms, and consumers. Unlike traditional logistics supply chains, catering distribution services have strong timeliness and consideration of delivery delay. Traditional distribution platforms usually group commodities by their origins or destinations and then transport each group with one logistics vehicle. However, for a logistics distribution area with limited commodities, the vehicle capacity cannot be fully utilized if one vehicle can only transport commodities with the same origin or destination. Therefore, a mixed-load strategy is proposed in which commodities with different origins or destinations in a distribution area could be transported by the same vehicle to improve vehicle capacity utilization. A mixed-load strategy would further cause delivery sequencing problems, leading to different delivery delays for customers. This study proposed an equity-oriented vehicle routing problem for food distribution services with timeliness requirements considering a mixed-loading strategy and vehicle capacity constraints. For the above problem, a multi-commodity flow optimization model was constructed for the equity-oriented vehicle routing problem and a mixed-load strategy based on a time-discretized space-time-state network representation. An augmented Lagrangian relaxation approach was utilized to reformulate the original model and thus effectively solve the proposed model. Furthermore, the augmented Lagrangian model was decomposed and linearized into a series of shortest path searching subproblems and iteratively solved by a dynamic programming algorithm using an alternating direction method of multipliers (ADMM)-based solution framework. Finally, the proposed model and solution approaches were tested on numerous networks.