System-Wide Optimization of Free-Floating Bike-Sharing for Urban Rail Stations: A demand prediction and scheduling approach

Abstract

Free-floating bike-sharing (FFBS) addresses the first/last mile challenges in urban rail transit (URT), while facing supply–demand imbalance problems owing to unrestricted bike parking. Previous research primarily equated actual bike usage with demand and focused on cost-efficiency, which overlooks unmet demand and system-wide optimization. This study proposes a comprehensive framework to optimize FFBS availability at URT stations, particularly during peak hours, through 1) demand prediction, 2) time-based scheduling, 3) priority scheduling strategy, and 4) system-wide optimization. The proposed method incorporates URT ridership as a pivotal feature to enhance the accuracy of bike-sharing demand prediction in URT transfer scenarios. To achieve bike-scheduling benefits, this study introduces a grid-based approach to convert ride data into predictive orders for bike scheduling, measuring time savings across transit modes. Additionally, a prioritization strategy for bike redistribution is designed based on the classification of bus routes around URT stations, ensuring a balanced integration of FFBS and other public transport modes. A multi-objective optimization model is designed to minimize operating costs and maximize passenger time savings, which is addressed with the NSGA-III algorithm. A numerical study using Shenzhen’s public transportation data reveals that prioritizing selected stations leads to a 19.4% greater average time savings per order compared to non-priority stations, along with a 7.60% reduction in total passenger travel time. This study more accurately reflects the actual demand, thereby achieving the supply–demand balance in URT-BBS transfers.