少量探針資料之平面道路交通資訊估測

Abstract

提供用路人各路段的即時交通資訊，並配合導航系統的使用，能有效的節省行車時間及能源消耗。由於具備GPS定位系統的車子逐漸普及，利用裝載GPS收發器的車子作為探針車取得即時交通資訊，並且回報給交通資訊中心來評估出目前的交通狀態成為可行的方式。一般來說，交通資訊中心都是將回報的資料直接採用平均及統計的方式來估測即時的交通狀態。但在少量回報資料時，會因為紅綠燈對探針車有不同的影響造成回報資料有較大差異性，並且導致無法準確地估測交通狀況。本研究基於少量回報的情況下，利用停走模式去降低探針資料被紅綠燈影響所造成的差異性，並讓交通資訊中心能有效的利用少量資料來準確地估測交通狀況。為了分析交通估測的準確性，我們以蜂窩浮動車數據 (Cellular Floating Vehicle Data, CFVD) 為真值並將估測的旅行時間與之比較。研究結果顯示，我們的演算法和CFVD之間的相關係數高於0.6，代表我們的方法和CFVD有高度相似的交通評結果，此外，平均絕對百分比誤差 (MAPE) 低於20%，顯示我們的停走模式可以提供精確的交通估測。

Providing real-time traffic information, such as the travel time of each road segment, to road users and on-board navigation systems can save travel time and reduce fuel consumption. With the increasing popularity of vehicles equipped with GPS receivers and wireless communication capability, the vehicles can be used as probes to collect the real-time traffic speed and travel time. In general, a traffic information center (TIC) uses the mean value of the reported data from probes and historic statistics to estimate real-time traffic conditions. However, when the probing data are sparse, the TIC may not be able to accurately estimate the traffic conditions of surface roads because traffic signals may cause delays of large variance on probe cars, and thus produce large variances among reported data. In this thesis, we propose a stop-and-go model in analyzing the probes’ GPS traces to reduce the variance in estimating traffic speeds with sparse probe data. To evaluate the accuracy of our estimation, we choose the travel time estimated from cellular floating vehicle data (CFVD) as our reference and compare our estimated travel time with that of CFVD. The results indicate that the correlation coefficient between our method and CFVD are higher than 0.6 which means that our algorithm and CFVD have similar traffic estimations. In addition, the mean absolute percentage errors (MAPE) are lower than 20% which indicates that our stop-and-go model can provide accurate traffic estimations.