JD.com Improves Delivery Networks by a Multi-Period Facility Location Model

Abstract

As delivery systems play an increasingly significant role in modern life, it is of strategical importance to design an efficient and adaptive delivery network. JD.com is the largest Business-to-Consumer (B2C) e-commerce company by revenue in China and has built the largest amount of logistics infrastructure nationwide. In this study, we developed an optimization model to automatically make annual plans of the delivery stations that minimize operational costs. We built an optimization algorithm by adopting a multi-period facility location model using Mixed Integer Linear Programming (MILP) method and proposing applicable techniques to solve the model. The results demonstrate that our model generates improved plans than manually planning in terms of both operational cost and delivery distance. The tool described in this paper is helping JD.com save delivery costs and gain operational benefits on designing delivery networks.

Keywords: dynamic facility location problem, delivery networks, mixed-integer linear programming