小汽車於市區道路各情景下之變換車道行為模式分析及比較

Abstract

近年來伴隨著人口快速增加及經濟成長，汽、機車數量大幅增加，造成交通問題日益嚴重，尤其對於市區道路之影響最為嚴重。為有效解決並改善相關交通問題，需透過微觀車流行為模式以了解車流之行為特性，而過去車流模擬軟體多以國外之汽車車流行為模式作為模擬運算之方法，然而國外車流模式多屬單純之小汽車車流所構建，鮮少考量國內較複雜之汽、車交織之混合車流之特性及影響。另外，相關研究大多僅針對定時定點之特定時空環境下收集車流資料並建立車輛行為模式，較少考量在不同道路幾何特性、尖離峰時段等差異之道路情景下是否對於車輛之行為模式產生影響，因此以單一特定時點車流資料建立車流行為模式之方式，也產生於不同情景下模式適用性之疑義。 為此，本研究透過拍攝車流影片，再以影像點取方式轉換為車輛移動軌跡之方式收集不同情景下之車流行為資料並採用自動化方式擷取相關影響變數和判斷車輛偏向行為，針對道路路段之小汽車車輛變換車道模式進行研究，採以個體選擇模式中之多項羅吉特模式為基礎，將車量維持向前直行、向左與向右變換車道視為三種不同之選擇方案。在考量與鄰車車輛相對距離、相對速度、車種等不同解釋變數使用下，建立四種小汽車於市區道路變換車道之變換車道模式並校估模式之解釋變數，期望透過不同解釋變數之選用與模式之改善及可找出適用於各情景下之選擇行為模式。最後再透過不同情景下收集之車流資料分別進行參數校估及檢定統計方法，以獲得各情景下解釋能力最佳之模式，並比較各情景的車流型態及分析影響模式差異性之因素，期望透過不同情景下車流資料庫之分析以及各情景下變換車道行為等比較，使變換車道模式達到能夠有效預測國內市區道路真實車流之行為。

With rapid growth of population and economic development, traffic congestion problem becomes more serious, especially for urban arterials. In order to model the traffic conditions accurately, it needs to understand the characteristics of traffic flow and behavior of vehicles for microscopic modeling. Most of the traffic simulation software available in the market is developed based on the traffic flow characteristics in foreign countries, and little attention has been paid to the effect of mixed traffic flow and urban arterials as in the cases in Taiwan. Previous studies calibrated their models using data from single location and time period, without considering the differences in behavior under various road types and traffic density. Therefore, the applicability of those models for the cases in Taiwan is in question. The objective of this study is to investigate the differences in lane-changing behavior of vehicles under various scenarios of road types and traffic conditions. We collect traffic data by taking videos of traffic movements at different locations for urban arterials in Taiwan. The videos are digitalized into trajectories and movements of vehicles, in which some explanatory variables and the vehicle lane-changing decisions are extracted. Lane-changing models are proposed and calibrated. The lane-changing behavior is formulated as a discrete choice problem, and the choices of a driver are remaining on the current lane, changing to the right lane, and changing to left lane. The models are calibrated using data from Taipei and Hsinchu, peak and off-peak periods. It is found that the calibrated results could be very different for different road types and traffic density.