請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/50154
標題: | 建立決策邏輯以預測被動警示牌面之成效 Establish decision logic to predict the performance of passive warning signs |
作者: | Chin-Hung Liu 劉錦鴻 |
指導教授: | 許添本(Tien-Pen Hsu) |
關鍵字: | 車聯網應用,複合安全指標,道路安全等級,資料探勘, Internet of vehicle,Composite safety index,Road safety level,Data mining, |
出版年 : | 2020 |
學位: | 碩士 |
摘要: | 相較傳統交通改善方法,車聯網(IoV)技術的應用有助於用路人提升對道路環境的感知能力,並提供即時資訊以輔助行駛的決策依據。本研究針對目前已投入於半開放場域的V2I(Vehicle-to-Infrastructure)機車安全警示系統,分析裝設警示系統前後道路安全的績效差異。本研究利用路側設施所蒐集之車輛紀錄資料作為判斷路段與路口分支安全績效之依據。相比肇事資料,蒐集車輛資料耗時較短,並能依此結果進一步推論具有較高潛在安全風險之研究地點。 本研究首先以平均偵測速率、加速擾度以及超/高速車輛比率三參數來描述各路口分支與路段之車流特徵與屬性,再透過主成分分析建立一複合安全指標,以K-means演算法將裝設牌面前後之複合指標數值進行分群,為各研究地點事前事後之安全程度分級。根據所求得之道路安全等級,以次序羅吉特模型建立顯著影響因子模型,判斷不同道路環境特性以及牌面設置等因子,是否對道路安全產生顯著影響。最後,同樣以道路安全等級數值作為模式預測值,建立CART決策樹模型,建立具有邏輯規則的結構來表示道路安全與影響因子之間的關係,作為評估牌面設置適用性之依據。 研究結果顯示,本研究之複合值與過去肇事資料分布具有中度相關,表示該指標具能夠以機車行駛行為評估其衍生之風險。並透過建立道路安全等級,發現裝設牌面多能提升研究地點之安全績效。對超速警示地點而言,根據次序型羅吉特與CART決策樹結果顯示,道路曲率半徑越大之地點道路安全等級越低,車道寬度、路面減速設施與是否裝設牌面均會對安全等級產生影響;鄰向警示地點則認為非路口支線方向具有較低之安全性,是否裝設牌面、道路曲率半徑、車道寬度以及牌面裝設位置等亦影響道路安全等級表現。 Application of IoV (Internet of Vehicles) technology has a significant advantage in raising drivers' perception of the road environment and providing real-time information to assist driving decisions. The objective of this study is to evaluate the differences in safety performance before and after the construction of a V2I (Vehicle-to-Infrastructure) motorcycle warning system which has been operated in a semi-open field. Motorcycle behavior records collected by roadside facilities are used as the basis for determining safety level of tested road sections and intersection branches. Vehicle dynamic records from the V2I system not only is more accessible than the accident data, but can also further identify the locations with higher road safety risks through analysis results. In this study, the three parameters of average detection speed, acceleration noise and proportion of speeding vehicles are established to describe the motorcycle traffic characteristics and attributes in each examined location and then be compounded to a unitary index by principle component analysis. After utilizing K-means algorithm to group composite index values, the road safety ranking of each site with two situations can both be classified. The property of ranking scores can help establish an order logit model to evaluate the significant impact factors on road safety. Finally, through the CART decision tree algorithm, this study constructs a logical rule structure to represent the relationship between road safety and impact factors, and also assess the applicability of the warning system. Results show that the composite index has a medium correlation with the distribution of the past accident data, indicating that it possesses sufficient interpretive ability to assess the road safety derived from driving behavior of motorcycles. Through the establishment of road safety levels, it was found that the installation of warning system can improve the safety performance of the tested site. For the speeding warning, according to ordered logit model and CART decision tree, curvature of the road, width of the lane, road deceleration facilities and whether installing of the warning sign will all affect the road safety level. For intersection warning, model strongly shows that branch direction has safer traffic environment. Other factors including whether installing the warning sign, curvature of the road, the width of the lane and the installed position of sign are also considered to influence the performance of road safety. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/50154 |
DOI: | 10.6342/NTU202002958 |
全文授權: | 有償授權 |
顯示於系所單位: | 土木工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
U0001-1108202015062100.pdf 目前未授權公開取用 | 3.9 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。