基於行動通訊網路訊號之市區道路即時交通資訊

Abstract

隨著市區交通擁塞越來越嚴重，提供道路用者即時的交通資訊，可以幫助他們避開擁塞的地區，進而節省旅行時間。相較於一般常見的車輛偵測器與裝載GPS接收器的探測車，因為手機普及率高而不需要額外的偵測裝置，利用無線通訊網路訊號推估即時交通資訊比較全面性且更具成本效益。這種透過追踪手機的位置來推測道路交通資訊的方法被稱為蜂窩浮動車數據(Cellular Floating Vehicle Data, CFVD)。CFVD目前已成功應用於高速公路，但是因為控制信號數量不足、難以辯識移動中的車輛及手機的定位準確度較低，所以在市區道路上成效有限。本研究提出了一個較新穎的CFVD演算法來估計市區主要幹道的交通資訊。首先，我們利用長時間的換手記錄統計找出相鄰的細胞區塊，進而確認移動站(Mobile stations, MSs)是否移動中。第二，我們使用行動通訊網路裡面的所有控制信令來估計交通資訊，並長期收集每一事件對(Event Pair, EP)的所有時間差(Time difference, TD)，將這些時間差做成累積分布函數(Cumulative distribution function, CDF)。然後，每一個TD可以參照相應的CDF轉換為百分比。最後，透過這些百分比可以融合不同EP的資訊產生自由流指數(Free Flow Index, FFI)及交通擁塞指數(Traffic Congestion Index)來反應交通狀況並估算旅行時間。為了驗證旅行時間的估計，這些估算出來的旅行時間將根據其值被分成五大類。由於實驗道路上沒有真值可以做比較，因此我們利用估計區間內的所有TD來驗證旅行時間的推估。研究結果顯示，FFI與TCI能夠有效地反應出交通狀況，而以TCI推估旅行時間有比較好的成效。

As more and more serious traffic congestion occurs in urban areas, providing real-time traffic information to road users can help them to avoid congested areas and thus save travel time. Compared with the conventional vehicle detectors and GPS-equipped probe cars, using control signaling of cellular networks to collect real-time traffic information has a wider coverage and is more cost-effective because mobile phones are ubiquitous and no additional devices needed on-road or on-vehicle. The approach has been referred to as Cellular Floating Vehicle Data (CFVD), which tracks mobile phones’ locations to estimate road traffic information. However, CFVD have been successfully used on freeways but fail on urban arterial roads because of insufficient control signals, difficulty in identifying moving mobile stations (MSs), and poor accuracy in locating an MS. In this thesis, we present a novel CFVD algorithm to estimate the traffic information of urban arterial roads. First, we use the long-time statistics of handovers to identify neighboring cells, and thus identify moving MSs. Second, we use all control signals of a cellular network to estimate traffic information, and collect the time differences of each event pair to generate the cumulative distribution function (CDF) of the time differences. Then, the time difference of an event pair can be converted to a percentile according to the corresponding CDF. Finally, we use these percentiles of different event pairs on the same road to generate Free Flow Index (FFI) and Traffic Condition Index (TCI), which can be used to estimate travel time. To verify the estimation, the estimated travel times are classified into five travel time bins according to their values. As there is no ground truth, we study the distribution of all time differences in each estimation bin to verify travel time estimation. The results indicate that FFI and TCI are useful indicators to estimate traffic condition and TCI is better than FFI in travel time estimation by a very small margin.