高速公路常因道路型態、交通變化或是駕駛者行為等因素，導致車流擾動情形發生。以收費站近似喇叭口特殊路段而言，在此前後漸變處，先是車道數增加、再則是車道數縮減情形；又由於收費所需，駕駛者在選擇合適付費孔道時，可能產生變換車道行為，造成車輛交織行為，上述兩者皆會影響車輛行經收費站旅行時間。本研究針對這些現象，將以國內一收費站為例，希望建立出駕駛者在收費站區域之行為模式及旅行時間預測模式。 本研究係以楊梅收費站南下車流為對象，希望透過實地拍攝方式，調查出收費站車流相關資料以作為模擬依據，進而建構出一套微觀車流模擬模式，以探討收費站此特殊路段下駕駛者行為。另一方面，藉由偵測器收集之資料，於不同流量下，利用倒傳遞網路法進行收費站區域旅行時間預測，並進一步分析預測模式之績效，最後期望提供精確之旅行時間。 經由反覆測試之結果得知，本研究所構建旅行時間預測模式，乃屬於「高精準預測」，於高速公路收費站區域旅行時間預測方面，可提供相關單位參考之用。 The traffic flow often disturbs on highway because of roadway patterns, traffic changes, or drivers’ behavior. Toll plazas are like bell, so the lanes add ahead and reduce backward. Because all the drivers must choose applicable paying lanes, they perhaps change lanes. It brings about vehicle interlacing. These phenomenons of the above which influence drivers’travel time. The research focus on this problem, then select one toll plaza to build up drivers’behavior and forecasting travel time models. This research is aimed at the southern direction traffic flow of Yang-mei toll plaza. Using monitoring and field search to investigate the data about traffic flow of toll plaza for simulation, and going a step further to construct simulation model with a microscopic view in order to discuss drivers’ behavior in toll plaza which a particular section of highway. On the other side, under different flow types, the research utilize back-propagation network algorithm to forecast travel time in toll plaza area, and to analyze this forecast modes. At last the result can prove exact travel time. Through repeatedly testing, the achievements of this forecast modes belongs to “high-precise”. The research can provide estimate for forecasting travel time in toll plaza area of highway.
