都市地區路網之動態資訊擷取與交通流參數演算研究

為達到先進運輸管理系統之運作，本研究提出一即時交通資訊擷取系統，規劃以計程車車隊為主體之探測車系統，結合路側偵測系統，以對於都市地區路網系統進行廣泛而完整之交通資訊擷取，並研擬系統運作下之資料庫系統設計。根據此一資料系統架構之下，考量資料項目之取得，本研究分別建構「路段旅行時間預測」和「動態旅次OD推估」之數學模式，並以廣義最小平方法和推廣卡爾曼濾波器進行模式之演算，對於路網系統之車流狀態進行預測，藉以支援相關動態交通控制、管理之決策。在模式驗證部分，本研究係透過Paramics軟體模擬一般化棋盤型路網系統下之交通車流，對路段旅行時間預測模式之預測結果進行分析，研究中主要藉由準確度、強健性和穩定性三個面向評估模式之預測能力。預測結果顯示，經由模式校估以及來源資料之處理過程，可得到優良之預測結果表現。對於動態旅次OD推估模式亦根據同樣的模擬結果進行試算之流程，將推估所得之旅次OD流量反應於路段流量上可得到良好之預測結果，是以評估模式之推估結果為合理。根據預測誤差，探討探測車回傳資料對於整體車流之代表性在模式預測準確度之影響，分析結果顯示，在車流中探測車佔有率達到5%以上，本研究所提出之路段旅行時間預測模式即可反應良好之預測能力。

關鍵字

交通資料擷取；動態旅次OD ；路段平均旅行時間；探測車

並列摘要

In order to establish an advanced traffic management system (ATMS), this study proposes a real-time traffic data acquisition system. The framework of system has its basis of taxi fleets as probe vehicles, and combines roadside detectors to collect traffic data from urban network extensively. According to the physical architecture, the study builds the mathematic models of “travel time prediction for road section” and “dynamic OD estimation”. The algorithms are based on generalized least squares (GLS) and extended Kalman filter (EKF) respectively. Through the prediction model, the traffic flows at various conditions are predicted to support decision-making of traffic control and management. To verify the models, for travel time prediction, this study analyzes the prediction results of grid network simulation from the Paramics, which is a traffic simulation package, and evaluates prediction capability by indices of precision, robustness and stability. The model proves a good prediction resulting in data calibration and processing. For the dynamic OD estimation, the study estimates traffic volume from predicted OD flows, which generates the mean error is within 10%. Accordingly, it concludes that the model is reasonable. Also, considering the representation of probe vehicles in the traffic flow, the model has good prediction capability, when the proportion of probe vehicles is above 5% of all vehicles.