- 會議論文
- OpenAccess
Travel Time Estimation and Prediction for Highways Using Current Flow Status and Historical Traffic Data Based on Internet of Vehicles Technologies
摘要
旅行時間的估計與預測往往決定了開車時的行車效率，尤其在高速公路上的行駛時間精準評估更是整體運輸效率的關鍵。因此，本篇論文利用高速公路ETC 單向資料傳輸之車聯網通訊技術所蒐集的資料，以高速公路之旅行時間估計與預測為主題，改善一般熟知的k-NN 分類法來預測擁有當日部分已知資料之路段旅行時間，並使用我們設計的斜率多重線性迴歸法來預測無任何已知資料之路段旅行時間，藉此幫助整體交通系統更易於管控，駕駛也得以擁有更好的行車品質。我們在擁有該日已知資料的狀況下，利用k-NN 法找出最相似的幾組歷史資料，觀察資料相依性找出最適當k 值以預測出精準的交通旅行時間;而在預測未來無當日任何資料的部分，我們將過去月份中相近日期時段與時間點的旅行時間資料分別做線性迴歸，並以它們的斜率變化量組合成適當的線性方程來預測未來的旅行時間，我們以國道高速公路局所開放存取的ETC 歷史資料實際驗證我們所提出的預測方法，實驗結果顯示所提出方法相較於現有方法可以大幅度地降低旅行時間預測誤差。
並列摘要
Travel time estimation and prediction are important for the driving efficiency in a long-distance travelling. The accurate travel time on the highway is the key to improve the efficiency of transportation systems. This paper focuses on travel time estimation and prediction for highways by enhancing the k-NN classification method to predict highway travel time when current flow status are known and by exploring the linear regression method to predict highway travel time based on historical traffic data when current flow status is unknown. The proposed estimation and prediction methods can be used to efficiently control the traffic management system and provide the best driving efficiency. With the collected data of current flow status, we use the enhanced k-NN method to determine the most similar historical data. The optimal value of k is obtained using traffic dependencies based on current and historical traffic data. When current flow status is not available, we use the slope-based linear regression method to predict travel time based on the linear relationship of the same consecutive weekday and weekend. To verify the feasibility and superiority of our proposed methods, we adopt the open ETC data of highways in Taiwan to evaluate the prediction accuracy of our approaches, which outperform existing methods and can significantly reduce the errors of travel time prediction.