高速公路動態車流動向推估模式之研究—以國道五號為例

Abstract

隨著現代科技的發展，應用相關技術與設備之智慧型運輸系統（Intelligent Transportation System, ITS），成為近年來交通運輸發展的重點，其中的先進交通管理系統（Advanced Traffic Management System, ATMS）更是ITS的核心。ATMS旨在藉由目前的交通資料，預測未來之交通狀況，並執行相對應之交通控制與管理措施。因此，若能掌握目前道路上車輛所欲前往之目的地，則更能夠正確預測未來之交通狀況，及早研擬策略，以減少交通問題。 鑑於上述背景，本研究以高速公路為目標，推估目前在高速公路主線上之車流，所可能離開之出口匝道，稱之為「動態車流動向」，構建動態車流動向推估模式。此推估模式以Lin and Chang（2007）之高速公路動態起迄研究為基礎，進一步延伸並將動態起迄轉換為動態車流動向。修改Lin and Chang模式中各車輛旅行時間服從某種分配型態之假設，改以偵測器之車速資料轉換為旅行時間；並且修改主線流率關係式與部分卡門濾波器求解流程。 本研究模式以國道五號雪山隧道北口至蘇澳端之資料作為實例運算，以主線流率比較運算之結果。驗證結果顯示，於運算結果中，各時階之MAPE皆不大於26%，大多數時階MAPE小於15%，表示推估結果多為「精確」或「良好」等級，可供未來ATMS作為運用。

As development of modern technology, Intelligent Transportation System (ITS) applied related techniques and facilities, become key point of recent progress of traffic. Especially, Advanced Traffic Management System (ATMS) is the core of ITS. ATMS aims to forecast future traffic by utilizing present traffic data, and implements corresponding traffic control and management methods. Therefore, if controllers know where the vehicles on the road are heading, they can forecast future traffic more accurate and make strategies to reduce traffic problems. Due to above-mentioned background, this study focuses on freeway and estimates “dynamic flow movements”, which means “which off-ramp that vehicles on freeway are heading to exit”, and constructs model to estimate it. This estimation model is extended by Lin and Chang’s model (Lin and Chang, 2007) which estimates dynamic freeway origin-destination. The dynamic flow movements estimation model transforms dynamic origin-destination into dynamic flow movements. Lin and Chang’s model assumes that travel time discrepancy among vehicles follows a certain distribution; this study modifies the assumption by transforming speed data into travel time discrepancy. This study also modifies mainline flow equations and parts of Kalman Filter process. This study uses national freeway no.5 as an example and compares mainline flow rates to validate the estimating results. Validation shows that estimation results MAPEs of each time-step are less than 26%, most MAPEs are less than 15%. It means most of the estimation results are “Highly Accurate” or “Good Forecast”, and this model is able to be applied to ATMS.