一個基於電信網路資料的智慧型運輸管理系統

Abstract

近年來隨著經濟的快速發展及科技的進步，許多傳統的運輸系統透過先進科技的協 助，獲得了有效的改善。這正是世界許多國家發展智慧型運輸系統(Intelligent Transportation System, ITS)所帶來的效益。即時交通資訊服務系統的推廣乃屬ITS重要 的一環。利用追蹤手機位置來偵測道路資訊的機制，這個機制不需要花費龐大的金額來 架設及維護額外的偵測裝置，而且幾乎每個人都有手機，因此我們以追蹤手機位置所得 到的交通資訊是非常全面的。本研究將提出指紋辨識演算法分析行動網路的位置服務事 件。針對已存在歷史紀錄之目標路段，運用迴歸分析建立車速、流量、交通密度三者的 關聯性，並透過網路服務(Location Service, LCS)定位演算法所取得之車速進行推論，取 得對應之流量和交通密度，以建立全面的交通資訊。對於未存在歷史紀錄之目標路段， 本研究提出運用手機通話(Call Arrival, CA)機率分析、週期性位置更新(Periodic Location Update, PLU)機率分析，於行動網路端進行CA和PLU的訊號收集，據此推論當下所對 應之交通密度；以及運用手機換手(Handover, HO)機率分析、一般性位置更新(Normal Location Update, NLU)機率分析，於行動網路端進行HO和NLU的訊號收集，據此推論 當下所對應之交通流量。並且運用車速、流量、交通密度三者的關聯性，估計當下的車 速資訊，並進行各種資料來源的資料融合，以產生更精確且全面的交通資訊。最後，本 研究採用雲端運算(Cloud Computing)針對大量的行動網路事件進行平行處理，並加入同 代像加密(Homomorphic Encryption)技術確保資訊安全，提供可靠且完整的交通資訊，予 以用路人參考。

The rise of fast economic growth and technology advance has led to improve the quality of traditional transport system. Intelligent Transportation System (ITS) has become more and more popular. The building and promotion of real-time traffic information system is an important part of ITS. The traffic information estimations from cellular network data are more immediately, cost-effective, and easy to deploy and maintain than traditional methods. In this study, we propose a novel speed estimation method using Location Service (LCS) events based on Fingerprint Positioning Algorithm (FPA). We then use history records and the Linear Regression Model (LRM) and to infer the equation between vehicle speed and traffic flow. ITS can use FPA, and LRM to provide the overall traffic information (i.e., vehicle speed, traffic flow, traffic density, and traffic condition). For missing history records, we propose a novel model to indicate the relation of Call Arrival (CA) rate, Periodic Location Update (PLU) rate, and traffic density; furthermore, this study proposes a novel model to indicate the relation of Handover (HO) rate, Normal Location Update (NLU) rate, and traffic flow. Then a data fusion algorithm is proposed to combine the estimated traffic density and flow from CA, PLU, HO, NLU rates for precise traffic information generation. Finally, we propose a modified homomorphic encryption algorithm and use cloud computing to analyze and process these large events from cellular network. This approach will be feasible to estimate the overall traffic information for ITS improvement.