Geo-fenced parking spaces identification for free-floating bicycle sharing system

Abstract

• Proposes a hybrid spatial clustering method to identify bicycle parking locations and capacities. • Studies the trade-off between coverage, average size and total number of parking facilities. • Presents an empirical study in Xiamen, China to test feasibility of developed method. Inconsiderate parking of free-floating bicycle sharing (FFBS) can lead to conflict between FFBS companies (due to their need for a viable business model) and the local government (due to concern on the negative public image made towards the city’s landscape). This is despite the importance of FFBS in providing a reliable first and last mile connection service. The provision of geo-fenced parking spaces, along with enforcement, is the most direct approach to deal with inconsiderate parking and has been adopted by government agencies. However, such parking spaces need to be designed properly to prevent a mismatch between parking location and capacity, and the true parking demand. In this paper, we proposed an easy-to-implement procedure, which integrates the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) method and the k-means clustering algorithm in identifying potential bicycle parking facility locations and capacities. Using the FFBS operations in Xiamen, China as a case study, real-time static bicycle location data were collected over regular time intervals throughout a 21-day period to calibrate and validate the hybrid-clustering model. Comparison between model results and existing bicycle parking lots showed that the parking lots determined by our hybrid-clustering model is more effective in terms of bicycle coverage than current practices. The selection of model parameters depends on the desired level of service and the designed available parking space per parking location. The influence of data scanning frequency on the results is studied and it is found that parking deployments model results are stable so long as the data scanning frequency is larger than 6 times per day. The proposed approach in this study will be valuable to the decision-makers as well as related shared mobility operators to determine the bicycle parking location and parking lot size at a strategic planning level.