在需求與供應不確定下自行車共享系統多站點間之供需媒合

Abstract

在公共自行車共享系統的營運期間，自行車的流動是動態的且分布不均勻的，為了避免使用者無車可借或無位可還，營運者必須在各租借站間以運補車來回調度自行車以達到供需平衡，重新平衡站點間的自行車數量對於維護服務質量至關重要，但是需要付出昂貴的成本。 自行車調度有兩種方式：靜態調度和動態調度。週期性的調度即是靜態的問題，動態調度能夠針對現場的實際情況，依環境的部分信息不斷更新調度策略，而動態調度的平衡問題具有需求不確定和供給不確定的特性。 本研究欲在共享經濟的概念下，針對動態調度改善自行車的調度問題，考量在短時間區段內，當需求不確定且供給不確定時，調度數量為何，研究中依照站點的現有車輛數，將站點區分為供應站點、常態站點和需求站點，在資源利用和客戶服務的兩個目標基礎上，分析不確定性需求與不確定性需求的站點配對問題，並求解非線性整數規劃，求得需求站點和供給站點的最適調度數量和配對，引入深度學習的概念，將站點資料輸入深度學習的模型，重複多組數值範例並從中找出調度數量與現有車輛、站點淨流量等站點資料之間的特徵且快速求解，預測最適調度數量，根據求得的調度數量提出建議。

During the operation of public bicycle sharing systems, the movement of bicycles is dynamic and uneven across bike stations and in different times of the day. In order to avoid stock-out and blocking, the operator must reallocate bikes back and forth between stations. To balance supply and demand, rebalancing the bikes between parking kiosks is critical to maintaining service quality, but it is costly. Bike rebalancing has two modes: periodical and dynamic. Periodical rebalancing is typically modelled as a static problem of vehicle routing. Dynamic rebalancing has the characteristics of both uncertain demand and uncertain supply. In this thesis, bike stations are categorized as surplus, normal and deficient stations based on their bike quantities and two thresholds. We present an analysis of the problem of matching uncertain supply with uncertain demand under the two objectives of resource utilization and customer service. We solve a non-linear integer programming model to obtain the optimal dispatch quantity and transfer between pair of station of supply and demand. Finally, we apply deep learning duplicate the optimal solution in a few numerical case.