Citation

Tian Z, Zhou J, Szeto WY, Tian L, Zhang W. Transp. Res. C Emerg. Technol. 2020; 112: 1-27.

Copyright

(Copyright © 2020, Elsevier Publishing)

DOI

10.1016/j.trc.2020.01.015

PMID

unavailable

Abstract

With the growing importance of bike-sharing systems, this paper designs a new framework to solve rebalancing problem. It contains two aspects: dynamic rebalancing within each station and static rebalancing among stations. Firstly, we give a new flow-type task window (F-window) by defining the consistency index of travelers. It is more suitable as a task window for rebalancing than time-type task window (T-window) based on three aspects analysis. Through three assumptions, the temporal-distribution learning model including task window and station storage configuration, are built to realize new dynamic rebalancing. The spatial-distribution learning method is introduced to divide management areas for static rebalancing. The empirical results show that F-window can better match the strong time-sensitive of demand fluctuation. Compared with traditional rebalancing needs hours, each rebalancing within a station can be completed within average 4 min. By setting the station storage configuration, it makes rebalancing in this paper meets the demand of 28.3 times the hourly rebalancing within one week. And the number of vehicles visiting stations has dropped below 20%.

Language: en

Keywords

Bike-sharing system; Community detection; Dynamic rebalancing; Flow-type task window; Inventory threshold; Spatial-temporal distribution learning; Static rebalancing