混合車流號誌化路口疏解時間預測模式之研究

Abstract

由於臺灣之機車持有率非常高，市區號誌化路口汽機車混合停等現象嚴重，為有效解決並改善交通問題，需要透過模擬軟體來了解混合車流車輛疏解之特性。然而，臺灣與國外的交通狀況有不少的差異，且國內外駕駛人的行為和車流型態也有許多不同之處，在車流疏解的研究上並不一定有考慮到臺灣交通特殊的部分，其發展的模擬軟體中之參數與模式無法直接沿用，且以往在藉由飽和流率求得路口容量僅是以一平均值去推估，並無法得知各週期之車流型態，所以用在模擬軟體上會忽略到許多因素。 因此，本研究在資料蒐集和建立軌跡資料庫後，以資料庫中數據去分析影響疏解之相關因素並撰寫程式將各變數描述和量化去探討混合車流疏解之情形。另外，將停等車隊分為3段：(1)機車車隊、(2)汽機車車隊、(3)汽車車隊，以統計之回歸分析方法，分別對在號誌化路口中之機車和汽車建構混合車流車輛疏解時間預測模式。最後，依照各段和各車種之車隊，分別得到共7個車輛疏解時間預測模式，並以MAPE和MAE等指標評估各模式和以往相關研究之預測準確性。

The holding rate of motorcycle in Taiwan is very high, and the phenomenon of cars and motorcycles stopping is chaotic at signalized intersections. For improving the traffic problems effectively, we need to understand the characteristic of mixed traffic flow. In the previous studies, the capacity of an approach of a signalized intersection is estimated by the saturation flow rates. However, the discharging behavior of the mixed traffic is also affected by the stopping pattern of automobiles and motorcycles, and the traffic patterns are not identical for every cycle due to the randomness of vehicle arrivals. These factors should be considered in determining the saturation flow rates and capacities of junctions. The objective of this study is to investigate the discharging behavior of vehicles in a mixed traffic at signalized intersections. Microscopic traffic data in the form of vehicle trajectory is collected, and the factors that could influence the discharge rate are analyzed. In the analysis, the queue platoon are divided into three sections: (1) motorcycle-only platoon, (2) motorcycle and car mixed platoon, and (3) car-only platoon. By adopting linear regression analysis, we propose a number of discharging time estimation models for different combinations of traffic mix. The result shows the discharging rate estimated by the proposed model has a good estimation performance.