Dynamic rebalancing strategies for dockless bike-sharing systems

Abstract

Bike-sharing systems have developed rapidly with the influence of the sharing economy, and many operational challenges have arisen. The bike rebalancing problem is one of the main challenges in bike-sharing systems. In this paper, we propose a framework to address the dynamic bike rebalancing problem in dockless bike-sharing systems by using trucks to relocate bikes to meet the time-varying demand at each location. We decompose the problem into two processes: dynamic clustering and bike relocation. For dynamic clustering, we propose an optimisation model to select cluster centroids and decide the number and coverage of clusters to maximise operational profit based on trip revenues and expected traversal costs between clusters. An Adaptive Large Neighbourhood Search (ALNS) algorithm is developed to solve this problem. Clusters with too many bikes would lead to bike piles and cause urban blight, while clusters with too few bikes may result in user dissatisfaction. To prevent such issues, in the bike relocation process, we construct vehicle routes with pickup and delivery for bike relocation between clusters. We test the framework using real data from Louisville, USA. We show that the proposed ALNS can efficiently solve large real-life instances and obtain high-quality solutions. Numerical experiments also indicate that the dynamic clustering model significantly increases average daily profit compared to static clustering benchmarks. We provide operators with several insights into the impact of clustering and relocation in bike-sharing systems.