基於車隊軌跡資料的計程車策略研究

Abstract

計程車載客服務，是一種常見的大眾運輸工具且大城市裡隨處可見，但計程車司機如何有效率的尋客，是影響到其收入的重要因子，而我們開發的路徑推薦系統期望提供司機適時適地尋客策略，期望在最短的時間內攬到新的客戶，縮短找尋客人的時間，增加個別計程車的使用率。根據大台北地區2015年6月約萬台計程車的歷史軌跡資料，分析客源及計程車數在不同時間與空間下的關係，並結合台灣運研所的電子圖資，以路段為單位進行排隊模型的分析及參數計算，包含: 載客率、等待乘客上車時間及路程時間，據以建立路網模型進而提出尋客繞徑的評估公式，並設計一個符合此公式heuristic的演算法，期望最短時間攬到乘客，並證明此演算法算出的路徑是最佳的。最後以我們提供之路徑策略與歷史資料統計的攬客時間相比，可以減少計程車司機49%巡迴攬客的時間，達到降低空車時間及提升使用率。我們也實作出一套系統，以網頁的方式呈現，依據不同時間及地點推薦一條路徑策略給司機並能顯示附近路段的攬客資訊，不只給予路徑建議也提供路段攬客資訊多面向的參考。

Taxi service is everywhere around the city. But allowing taxi drivers to pick up passengers more efficiently is one of the most important things. We hope to develop a system which can provide a better route for taxi drivers, and pick up passengers more efficiently. It can also reduce the time to find the next passenger, and increase the usage rate. We analyze where to pick up the passengers at different times and places through about ten thousand historical tracks of taxis in June of 2015. Besides, we also use the geographic information by Institute of Transportation, MOTC, analyzing and calculating according to each unit road segment, which includes the rate of picking up passengers, the time spent waiting for a taxi and the travel time between connected road segments. Constructing a road network based on the predicted formula derived from the route for searching for passengers, and design the algorithm for heuristic formula. We hope to use the shortest time to pick up the passenger. Finally, we compare the routes of our results and the historical data, it shows a 49% reduce of time of free taxis. We also developed a system, it can recommend the route for each taxi driver and show road information where to pick up the customer.