SDVRP-Based Reposition Routing in Bike-Sharing System

摘要

Bike-sharing systems have recently been widely implemented. Despite providing green transportation method and a healthy lifestyle, bike-sharing systems also poses problems for system operators: In order to meet the public's demand as much as possible, operators must use multiple trucks to relocate new bikes and repaired bikes from the depot to different stations. Then, the route to minimize the cost for the delivery trucks becomes a serious problem. To address this issue, we first formulate the problem into a split delivery vehicle routing problem (SDVRP) since every station's demand can satisfied by multiple trucks, and use the K-means algorithm to cluster stations. In general, K-means is used to cluster the nearest points without constraint. In this real-world constraint problem, the sum of zones' demands must be smaller than total truck capacity. Therefore, we transform the SDVRP into a traveling salesman problem (TSP) by using a constrainted K-means algorithm to cluster stations with the demand constraint. Finally, according to the context, we use a genetic algorithm to solve the TSP. The Evaluation considers four real-world open datasets from bike-sharing systems and shows that our method can solve this problem effectively.