應用車輛辨識系統提昇起迄旅次矩陣推估之研究

Abstract

起迄旅次矩陣（Origin-Destination matrices，O-D matrices）係交通管理政策研擬、交通設施規劃，以及交通資訊提供之重要參考資訊，其又可分為動態及靜態兩大類。然在未實施匝道閉闔式收費前，高速公路起迄交通量僅能透過龐大之路邊或家戶人工調查方式加以取得。而一般非封閉式平面道路，更難以取得此項資訊。因此，許多研究乃轉而利用路段所測得之流量資訊反推起迄交通量。然而，利用路段流量反推起迄交通量之問題仍具有相當難度，其主要原因乃是因為路段流量方程式之數量大幅少於所欲求得之起迄對交通流量變數之數量，導致該問題形成一低度確定問題，無法求得單一確定解。近年來由於智慧型運輸系統（intelligent transportation system, ITS）之迅速發展，許多有助於起迄旅次矩陣推估之資訊也日漸普及。其中，用於自動收費及旅行速率推估之車輛辨識系統（automatic vehicle identification, AVI）即可利用車輛裝置之特殊裝置（如tag、OBU或GPS）或透過車牌辨識（license plate recognition, LPR）技術「追蹤」某一輛車，以提供其部份路徑（partial trail）資訊，用於進一步提昇起迄交通量矩陣之推估精確度。本研究即旨在探討在考量推估動態或靜態起迄旅次矩陣之精確度最大化及用路人隱私權侵犯最小化之目標下，利用雙層數學規劃模式（上層決定AVI區位，下層推估起迄矩陣），用以決定最佳AVI設置數量與區位。

The origin-destination (O-D) matrix is essential for proposing traffic management policy, planning transportation facilities and providing real-time traveler information, which can further be divided into two categories: static and dynamic O-D matrices. However, for a freeway system without ramp-closed tolling system or automatic vehicle tracking system, a large-scale O-D roadside or household survey might be the only way to obtain such information. Therefore, numerous researchers propose and solve the problem of estimating an O-D matrix from traffic counts on road links. This problem is challenging because the number of links with traffic counts is usually much smaller than the number of O-D trips to be estimated in the matrix. Of course, it is impossible to identify a unique optimal O-D matrix from link counts alone in most of cases. Recently, with the rapid growth of automatic vehicle identification (AVI) techniques, many facilities or devices can be used to largely enhance the estimation of O-D matrix. Some of which are commonly used to electronic toll collection or estimation of travel time by means of tag, OBU or GPS equipped in the vehicles or through the technique of license plate recognition (LPR), to track a specific vehicle. With the partial tail information of the vehicle provided, the estimation of O-D matrix can be greatly enhanced. However, wide installation of AVI may seriously invade the privacy of road users and consume a great deal of government budget. Therefore, this study aims to determine the optimal number and locations of AVI for the purpose of enhancing the estimation of an O-D matrix at minimal privacy invaded and installation cost through a bi-level mathematic programming model. The upper level is to determine the optimal locations of AVI and the lower level is to solve the O-D matrices.