公車到達時間估測方法之研究

Abstract

傳統上在公車實際運行時，往往容易受到交通壅塞、突發交通事故或天候不佳等因素，使得公車無法按照時刻表上的時間到站，造成乘客不便。近年來，隨著資通訊與電機電子技術的高度整合，智慧型運輸系統(Intelligent Transportation System, ITS)的科技已日漸成熟，在現今的公車運輸系統中，已能夠透過公車上的車載機(On-Board Unit, OBU)提供可靠且即時的交通資訊，藉此進行準確的公車到達時間預估，提昇乘客對於運輸系統之滿意度。 為了善用即時之交通資訊進行公車到達時間之預估，本論文基於臺中市交通局所發展之公車動態資訊系統，提出一種對公車即時動態資訊之巨量資料擷取與分析方法，建立公車各路線之行駛路徑函數與行駛時間歷史資料庫，並提出基於行駛路徑函數之公車到達時間預測方法與臺中市交通局與中華電信專利之估測方法進行評估比較。有鑑於現行估測方法對公車停滯時間的考量不足，本論文進一步以公車運輸交通模型(Bus Transportation Traffic Model, BTTM)模擬公車在行駛路段上之交通車流與乘客上下車行為，並使用非線性卡曼濾波器來預測系統狀態。綜合上述，本論文在公車即時動態資訊提供下，完成一套完善且可行的公車到達時間線上預估方法。

Traditionally, the actual arrival time of a running bus might be later than that displayed on the bus timetable due to taffic jams, car accidents or the bad weather condition which may yield inconvenient for passengers. Recently, the technologies of information, communication, electrical and electronics are highly integrated, so that the solution of ITS (Intelligent Transportion System) is getting well-developed. In the current public transportation systems, it has great chance to obtain the reliable and real-time traffic information via the OBU (On-Board Unit) on each bus, in order to obtain ETA (Estimated Time of Arrival) of each bus and satisfy the requirements of passengers. Above all, in this thesis, based on the Taichung City Dynamic Bus Information System, a novel scheme of big data acquiring and analysis is proposed to facilitate the usage of the real-time traffic information. The accurate route-path function and the traveling-time historical database of each bus route will be built up properly to assist the construction of arrival time estimation method. The comparison and evaluation among the proposed method and the two known estimation methods used by Taichung City and Chung-Hwa Telecom are also going to highlight contributions. Moreover, a novel traffic model BTTM (Bus Transportation Traffic Model) is proposed to simulate the traffic flow and the behavior of passengers, and also the non-linear Kalman filter is also used to predict the system states such as average speed and dwell time.