機車於市區道路中車道規範之研究

Abstract

台灣的機車持有率為世界第一，2015年平均每千人有582輛機車，然而機車於道路中的駕駛行為與小汽車有所不同，汽車於道路中行駛時須遵守車道規範的限制，亦有法律規範汽車須行駛於車道中間，不得跨越車道線行駛，然而機車因體積小，可在車道中任意選擇位置行駛，稱機車無車道規範的限制，因此，機車會有其獨特的駕駛行為，像是超車、鑽車、蛇行等等。市面上一般微觀車流模擬軟體為歐美國家所開發，而歐美國家的車流環境主要以小汽車為主，與國內混合車流的背景不同，無法充分反映機車在道路中的行為特性。本研究將探討機車在市區道路中是否符合車道規範限制，探討機車在路段中的橫向位置以及超車的行為特性，並發展機車在道路中的橫向位置選擇模式，並將模式應用於微觀車流模擬軟體中，期望能提升模擬軟體的準確度，另外，本研究亦利用微觀車流模擬軟體，探討有無設置禁行機車道對車流的影響。 微觀車流研究常受限於資料取得不易，本研究利用無人機進行空中拍攝，可拍攝完整路段的車流狀況，調查地點包含不同車道數路段，且錄製不同車流密度的車流影像，並利用車流軌跡軟體(Lee et al., 2008a)擷取車輛於路段中的軌跡，建立車輛軌跡資料庫，以此資料庫進行機車橫向位置的分析。 經由實際資料發現，汽車於路段間均維持車道規範限制，多行駛於車道線與車道線中間，而機車在路段的橫向位置選擇的分布橫跨車道線，呈現一個三角形分布，因此本研究針對機車於路段起始位置的橫向位置分佈進行分析，校估不同車道數、不同密度下的機車橫向位置分佈模式，且藉由回歸模式的變數最小化模式的RMSE值，並利用MAPE、R^2等評估指標判斷此模式表現是否良好，研究發現機車進入路段時的橫向分佈會受車流密度、禁行機車道、路側設施等影響，最後亦針對模式變數進行敏感度分析，確認本研究模式的穩定性。 本研究探討現有車流模擬軟體VISSIM與BikeSim對於機車橫向位置的彈性，並比較模擬結果與實際資料的差異，並將機車橫向位置分佈模式加入微觀模擬軟體BikeSim中進行模擬，探討加入模式後模擬結果是否更符合真實車流情況，模擬結果顯示加入研究模式後能提升模擬軟體的準確度。最後，亦模擬開放禁行機車道對於車流的影響，結果顯示開放禁行機車道會造成機車在內側車道的車速上升，但對於汽車較無明顯的影響。

The holding rate of motorcycles in Taiwan is the highest in the world. As the amount of motorcycles traveling on the road, the behavior of motorcycles are extremely different from cars. On the roads, cars should comply with the lane discipline to drive in the middle of the lane. Since the size of motorcycle is smaller than car, motorcyclist can choose any position to drive. On the other hand, the driving behaviors of motorcycles are not limited in lane discipline. However, the traffic simulation software developed by European and American countries is mainly based on the driving behaviors of cars. The simulated results are unable to present the non-lane-based movement of motorcycles in Taiwan. The aim of this study is to understand the lane discipline of motorcycles in mixed traffic by empirical analysis, and develop a selection models of lateral position of motorcycles. Besides, we integrate the model into a traffic simulation software to enhance the accuracy of the simulated results, and use simulation software to compare the effect of setting motorcycle prohibited lane. The earlier researches of microscopic traffic flow were usually limited in data collection. In this study, we collected traffic data using aerial videography by a multicopter, which could be used to record the video of the traffic on the road. The survey locations include different numbers of lanes and density. Then, we used the trajectory extractor software to collect the trajectories of vehicles, and built up the database for analysis. By real data, we observed that drivers of cars usually drove in the middle of lane. However, the distribution of lateral position of motorcycles on the road was across the road, and presented a triangular distribution. Therefore, this study aimed the lateral position of motorcycles on the starting position of road to analysis, and calibrated the model in different numbers of lanes and density. The variables of model were calibrated by regression analysis, and we used RMSE, MAPE, and R^2 to determine the performance of model. Furthermore, this study explore the flexibility in modeling lateral shifts in two kinds of microscopic traffic simulation software, VISSIM and BikeSim, and comparing the differences between the simulated results and real data. Besides, we integrated the model of lateral position of motorcycles into BikeSim, and the simulation results confirm that the accuracy would increase. Moreover, we use BikeSim to compare the effects of setting motorcycle prohibited lane, and the results show that there are no significant effects on cars, but the speed of motorcycle driving on the inner lane would increase.