Researching the Optimization of Bike-Sharing Dispatching Route Based on the GA-ALNS Algorithm

Abstract

Bike-sharing systems (BSS) are essential to promoting sustainable urban mobility. However, their operational efficiency is often hindered by the imbalanced distribution of bicycles across stations, leading to user dissatisfaction and increased operational costs. Addressing this imbalance requires optimizing the routing of rebalancing vehicles—a complex task typically modeled as a vehicle routing problem with simultaneous pickup and delivery (VRPSPD), an NP-hard problem. Therefore, this study proposes a hybrid algorithm combining genetic algorithm (GA) and adaptive large neighborhood search (ALNS), collectively termed GA-ALNS. A mathematical model was formulated to minimize the total cost, encompassing travel distances and penalties for unmet demand. Within the GA-ALNS framework, GA executes a global exploration to identify promising regions of the search space, while the ALNS—enhanced by an adaptive mechanism that dynamically selects destroy and repair operators—performs intensive local refinements. Computational experiments based on real-world data demonstrate that the proposed GA-ALNS algorithm significantly outperforms benchmark algorithms, including standalone GA and ALNS. The hybrid method achieved an average reduction of 7.5% in total scheduling costs and successfully balanced over 95% of the target stations compared to the best-performing benchmark. Furthermore, the algorithm exhibited robust performance and rapid convergence across diverse problem instances. These findings provide a valuable reference for addressing similar logistics optimization problems and present a practical solution for BSS operators, contributing to lower operational costs and improved service quality.