共享單車系統之即時需求預測與運補決策

Abstract

近年來共享單車系統（Bike-sharing system）的普及使大眾的使用率大幅提升，目前此系統在人口密集的大城市中面臨的重大問題是使用者常常遇到無車可借或無位可還的窘境。為了改善此問題，YouBike公司現行的做法是派遣運補車到附近的車站進行運補，但是目前運補的依據和準則仍有許多缺陷，例如在尖峰時間使用者的流量較大，等到車站無車可借或無位可還時再派車進行運補往往為時過晚。本研究提出一個共享單車使用者的租借與歸環機率模型，並利用實際的單車數量資料變化來進行使用者流量的估計，最後結合排隊理論建構出運補車的最佳運補決策，藉此減少無車可借或無位可還的情況，並提出一個分群演算法配置最佳的運補車據點位置以提升運補效率。本研究將以台北市YouBike 2.0的車站資料為例，以模擬的方式來驗證所提研究方法的成效。

In recent years, the widespread adoption of bike-sharing systems (BSS) has led to a significant increase in their usage. However, a major issue faced by such systems in densely populated cities is the frequent unavailability of bikes or docking spaces for users. To address this problem, YouBike currently dispatches rebalancing trucks to nearby stations to redistribute bikes. However, the current rebalancing criteria and policies have many shortcomings. For example, during peak hours, user traffic increases significantly, and waiting until a station runs out of bikes or docking spaces to dispatch trucks is often too late. To overcome the shortcomings, this study proposes a rental and return probability model for users, estimating the users’ flow based on actual bike count data. Furthermore, by integrating queueing theory, we construct an optimal dispatching policy for rebalancing trucks to reduce situations where users cannot find available bikes or docking spaces and proposes a clustering algorithm to determine the optimal locations for rebalancing truck depot to improve rebalancing efficiency. The study will use data from Taipei's YouBike 2.0 stations and employ simulations to verify the effectiveness of the proposed methods.