306 / 2024-04-24 11:24:43

A data-driven approach to the city last-mile delivery problem towards the application of shared delivery terminals

Last-mile delivery; Shared delivery terminals; Data-driven approach; Predictive optimization; Support vector regression; Adaptive large neighborhood search

The application of shared delivery terminals is a promising trend in the development of last-mile delivery in city logistics, as it can effectively improve the delivery efficiency of couriers and relax the time window restrictions for customers to pick up their parcels. In this paper, we study a vehicle routing problem (VRP) for the application of shared delivery terminals in last-mile delivery. In practical delivery scenarios, the random storage and retrieval behaviors of customers can affect the usage of shared delivery terminals and lead to inevitable uncertainty regarding the available capacity of shared delivery terminals, thus increasing the complexity of last-mile delivery. To address this issue, we propose a VRP with stochastic terminal capacity (VRPSTC) and design a data-driven predictive optimization approach by collecting first-hand usage data of shared delivery terminals, forecasting the available capacity, and optimizing the operational delivery schedule in practice. Numerical experiments show that the proposed data-driven approach can effectively solve the proposed VRPSTC, and contribute to around 17%-20% reduction of total delivery cost in comparison with the traditional stochastic optimization. It is expected that the proposed VRPSTC can enrich the context of last-mile delivery in terms of both theoretical research and practical industrial application.