Machine learning-based optimisation in a two-echelon logistics network for the dry port operation in China

Abstract

ABSTRACT Efforts of creating economic recovery after the COVID-19 pandemic stipulate international logistics demands in the countries and regions affected by China's Belt and Road initiative. Considering the increasing number of small and middle-sized enterprises, there is a great challenge to make transportation plans for the hub port, dry ports, and related enterprises. We investigate a two-echelon vehicle routing problem with simultaneous pickups and deliveries. On one hand, freights are transported from a central depot to multiple satellites, then distributed from the satellites to customers. On the other hand, freights collected from the customers will be loaded at the satellites, then transported back to the depot. We model the problem as mixed integer programming and propose a machine learning-based hybrid algorithm to solve the problem. Our hybrid algorithm comprises a K-Nearest Neighbour algorithm and an Adaptive Large Neighbourhood Search heuristic. We apply our modelling and solution approach to a real case based on the online digital platform ‘Inland Port Cloud Wharf’ in China, which matches international demands of commodities with domestic supplies. Our computational experiments based on a practical case study show the efficiency of our KNN_ALNS algorithm in optimising networks with multimodal coordination and addressing real-world logistics complexities.