- 學位論文
使用車輛偵測器和自動車輛辨識之資料預測高速公路旅行時間
Travel Time Prediction for Freeway Using Vehicle Detector and Automatic Vehicle Identification Data
摘要
台灣的運輸路網發達，部分時段交通量大造成交通擁塞，如能提供準確的交通資訊給用路人，或能幫助用路人避開擁塞時間及路段，也能降低路段的擁塞程度。高速公路上最重要的交通資訊為旅行時間，本研究的目的是利用國道高速公路的車輛偵測器(Vehicle Detector, VD)及自動車輛辨識(Automatic Vehicle Identification, AVI )作為資料來源，建立旅行時間的預測模型。使用VD具有即時集交通狀況的特性，透過模式辨認(Pattern Recognition)的方法與VD旅行時間歷史資料庫比對，找出最相符的日期與時間點，再以相對應的AVI旅行時間歷史資料進行預測。本研究使用模式辨認中的k-NN(k-Nearest Neighbor)法進行交通狀況的比對。為了增加比對的準確性，本研究在k-NN模式中將比對時間長度( )列為校估參數並且對k設定條件限制。將比對時間長度列為參數，可找出更適合路段的比對時間長度，來提升預測準確性。在偵測器不穩定時，所集的交通資訊與現實交通狀況不符合，因此對k限制同一天比對成功的筆數限制，可以避免過多的資訊來自同一天，減少偵測器不穩定所造成的誤差。最後以實際案例探討，探討加入AVI資料是改善預測準確性的情況。
並列摘要
Accurate traffic information can help road users to avoid traffic congestion and reduce delay. The objective of this study is to develop a travel time prediction model for freeway using data from Vehicle Detector (VD) and Automatic Vehicle Identification (AVI). VD data can be collected and used in forecasting real-time, but usually suffers from inaccuracy when the traffic is congested. On the other hand, accurate travel time information can be derived from AVI data, but the collection AVI data and license plate matching have a delay and cannot be used in real-time prediction. We combine the two data sources and establish a historical database. With pattern recognition technique, we can identify the most similar traffic pattern in the database using real-time VD data for prediction. Based on the k-nearest neighborhood (k-NN) method, an modified approach is proposed with additional model parameters. The results show that the prediction model with combined VD and AVI database is more accurate than the model with only VD data.