有GPS資訊提供下之公車旅行時間之研究

Abstract

由於智慧型運輸系統(Intelligent Transportation Systems, ITS)中的先進大眾運輸系統(Advanced Public Transportation System, APTS)之電子技術能有效改善大眾運輸系統，進而吸引更多人搭乘大眾運輸系統，使得交通壅塞得以有效減輕。又先進大眾運輸系統中的即時車輛到站資訊可以改善大眾運輸系統中的派遣任務，並且提高乘客搭乘大眾運輸運具意願。因此旅行時間的預估就變的非常重要。

由於過去研究對於國道旅行時間預估多以佈設大量偵測器收集所需相關資料，然而要在市區佈設大量偵測器需耗費較大成本，加上近年來車輛運用全球定位系統(Global Positioning System, GPS)以達車輛自動定位漸漸普及，因此本研究將以市區公車裝配之GPS回傳的即時資料為本研究主要資料來源。

本研究將公車旅行時間切割為公車運行時間和公車停等時間。其停等時間又分為兩部分，第一部份為公車於交叉路口停等號誌之停等時間，第二部份為公車在停靠站牌載客上下車所發生之停等時間，本研究在停等時間僅針對第一部份停等時間預估。

本研究主要透過車輛歷史車速資料預估車輛運行時間，並自行發展預估模式來預估停等時間。其中在停等時間之預估，將以改變點分析將歷史資料庫依車速高低之不同型態切割成數個不同時段，而在不同車速之時段，本研究將以不同預估模式來預估公車於交叉路口停等號誌之停等時間。

最後本研究將以實際營運之公車業者為測試對象，並針對各站之預估到站時間、旅行時間和預估停等時間分析其準確度。

Attracting more people to take advantage of public mass transit systems, the electronic technologies in Advanced Public Transportation Systems (APTS) of ITS can gradually ease traffic jams. Among all these, the real-time vehicle arriving information helps most on dispatching jobs, thus improves the will that people take public mass transit. Therefore, the importance of travel time prediction cannot be over-emphasized. In the past, most of the relative researches all gather their data through a lot of detectors which are needed to be installed, and that costs a big deal whether in time or money. But nowadays, as GPS (Global Position Systems) is generally popular in positioning vehicles, in this study we choose the real-time data by buses equipped with GPS to be our major data source. We divide bus travel time into “bus running time” and “bus waiting time”. The latter is composed of two parts: the first part is the waiting time that buses wait for signal change in intersections; the second part is the waiting time that passengers board on the buses at bus stops. In this study, we concern only the first part. The methodology to estimate travel/waiting time in this study is through historical vehicle speed data and the prediction model we construct. In the prediction of waiting time, we classified the database by vehicle speed types into several periods; in each period, different prediction models are used for predicting the waiting time that buses wait for signal changes in intersections. Also, we take real data as samples. In every bus stop, different prediction time (arrival time, travel time and the waiting time) are respectively analyzed to test the accuracy of the model