構建固定式車輛偵測器遺失值之快速插補模式

機構典藏 > College of Business and Management > Graduate Institute & Department of Transportation Management > Thesis > Item 987654321/105464

Please use this identifier to cite or link to this item: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/105464

Title:	構建固定式車輛偵測器遺失值之快速插補模式
Other Titles:	Development of a quick response interpolation model for missing data of vehicle detectors
Authors:	陳宛靜;Chen, Wan-Ching
Contributors:	淡江大學運輸管理學系碩士班董啟崇;Tong, Chee-Chung
Keywords:	遺失值;車輛偵測器;轉換函數;插補模式;Missing Value;Vehicle Detectors;transfer function;Interpolation Model
Date:	2015
Issue Date:	2016-01-22 14:56:35 (UTC+8)
Abstract:	今日道路運輸系統逐步運用ITS技術作為管理工具，現階段掌握路段即時狀態以固定式車輛偵測器資料為主。惟在實務運作中，偶有面臨車輛偵測器本身故障或傳輸過程失常所出現資料遺失情況，導致路網交通狀態的判定結果產生偏誤。因此，如何以即時、有效方式解決遺失值問題，提供更完整、可靠路段資料供交通管理者快速反應路網狀態，即是本研究建立資料遺失插補模式之研究重點。回顧過去國內外文獻針對交通資料遺失插補研究，雖已提出相當多元處理方法，且伴隨其特定公式與推導理論為佐證，但具有不同之特性與限制，並無廣泛最佳模式之共識。本研究歸納分析值得重視之焦點在於 (1)甚多模式普遍基於複雜的理論，且推算過程相對耗時費力，或(2)運用近似黑盒子軟體作為運算輔助工具，對模式與參數的意義與掌握稍顯薄弱，(3)模式參數與運用之空間移轉性問題亦少深入探究，因此對於實務上追求模式簡便、快速、實用之運算特性仍有些許落差。本研究將以符合車流理論之時空推移關係為基礎，以簡單、快速、易於操作為導向，透過能描述序列資料之間具時空相依關係的多元輸入轉換函數模式，作為建構具有穩定參數值和較廣泛適用之快速插補模型，在不同但相似道路類型下之任一不特定封閉線性路段，插補出於可接受範圍內之補值結果，因此，除了運用過去文獻常使用之鄰近偵測資料建構模式外，本研究並學習類比車流理論中構建跟車(Car-Following)行為之通用汽車模型(GM Model)發展過程建立階段模型，界定先以即時取得的鄰近速率為錨定值，對應於上下游間之即時速率變動量來反應中間待插捕速率變化之基本模式，另加入偵測器佈設間距調整參數和車道增縮調整參數來鬆綁插補模型，以求模式推估效果能依模式演進過程提升插補準確率。此外，本研究考量車輛偵測器資料可能有長期或短期遺失情況，分別構建長期和短期遺失值之最適插補模型。本研究選定國道一號北區和南區局部路段為範圍，並依其車流速率序列狀態歸類為不穩定和穩定車流速率資料分別構建與校估模式。經自我和交叉驗證程序後顯示模型插補估算表現大致皆可達高精準度之績效 (MAPE值<10%)，包括在高速公路非封閉線性路段或封閉公路環境之平常日的不同車流狀態下(含過渡與壅塞)之長期和短期遺失值插補與在特殊連續假期(如過年一週連假)之極端尖離峰車流狀態下之長期遺失值的插補估算等。整體而言，現階段模型插補表現確實達到本研究初始期望合乎實務訴求之簡單、快速、易於操作之運算特性，而模式應用範圍的彈性也符合達成廣泛運用之目的，亦可提供趨向發展標準化模式(unified model)之參考(基石)。 Technologies of Intelligent Transportation System (ITS) have been adopted gradually in road traffic management system, where data collected through vehicle detectors (VD) deployed widely in roadway network are the main resources for the management authorities to recognize real time traffic condition along road segments. However, judgment of traffic link performance has been seriously compromised by missing data issues encountered from either functional impairment of detector devices and/or data transmission process. As such, missing data interpolation is the targeted issue for this study to establish an effective approach to generate complete and reliable link performance data. Various techniques have been developed for missing traffic data interpolations based on different theories and formula with certain level of success but been subjected to some sorts of limitations. As such, there was no one technique or method regarded as the best model. The concerns of this thesis focused on the following issues for practical applications: (1) many techniques employ very complex procedure (2) some rely on black-box type of operations, which may be lack of tractability (3) stability and transfer ability of model parameters were seldom fulfilled. It is therefore the major perspective of this study to develop a practical method with relatively simple and easy operations. Based on spatial-temporal characteristic of traffic flow, transfer function technique was selected to develop the so-called “quick response” missing speed data interpolation model applicable to varieties of road segments. Model specifications and development were inspired by the generations of General Models on the car-following traffic flow theory. Either the most adjacent up-stream or down-stream attainable VD data associated with the missing segment were used as anchored values in junction with the differences values between them accounted as major exploratory variables to establish the missing values in-between. Adjusted factors were further specified to account for changes in number of lanes and relative distance of missing data VD to the anchored VD. Data from the selected regions of National Freeways and highways were employed for model estimations and validations. Self-validation and cross validation were implemented. The results have shown that highly accurate estimates were obtained for missing values of speed (with MAPE values being less than 10%) in most cases on both short-term and long-term periods. The findings from this study may provide a foundation towards development of the unified model in missing traffic data interpolation operation.
Appears in Collections:	[Graduate Institute & Department of Transportation Management] Thesis

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	140	View/Open

Loading...