Dynamic Flow Distribution Prediction for Urban Dockless E-Scooter Sharing Reconfiguration

Abstract

Thanks to recent progresses in mobile payment, IoT, electric motors, batteries and location-based services, Dockless E-scooter Sharing (DES) has become a popular means of last-mile commute for a growing number of (smart) cities. As e-scooters are getting deployed dynamically and flexibly across city regions that expand and/or shrink, with subsequent social, commercial and environmental evaluation, accurate prediction of the distribution of e-scooters given reconfigured regions becomes essential for the city planners and service providers. To meet this need, we propose GCScoot, a novel dynamic flow distribution prediction for reconfiguring urban DES systems. Based on the real-world datasets with reconfiguration, we analyze the mobility features of the e-scooter distribution and flow dynamics for the data-driven designs. To adapt to dynamic reconfiguration of DES deployment, we propose a novel spatio-temporal graph capsule neural network within GCScoot to predict the future dockless e-scooter flows given the reconfigured regions. GCScoot preprocesses the historical spatial e-scooter distributions into flow graph structures, where discretized city regions are considered as nodes and their mutual flows as edges. Given data-driven designs regarding distance, ride flows and region connectivity, the dynamic region-to-region correlations embedded within the temporal flow graphs are captured through the graph capsule neural network which accurately predicts the DES flows. We have conducted extensive empirical studies upon three different e-scooter datasets (>2.8 million rides in total) in populous US cities including Austin TX, Louisville KY and Minneapolis MN. The evaluation results have corroborated the accuracy and effectiveness of GCScoot in predicting dynamic distribution of dockless e-scooters' mobility.