基於GPS軌跡估計紅綠燈時相

Abstract

隨著市區道路使用者大幅增加，交通壅塞會導致使用者增加旅行時間，我們發現紅綠燈的時相對於旅行時間有很大的影響提供道路用者交通指標以及紅綠燈時相資訊，可以幫助使用者們避開壅塞的路段以減少旅行時間。由於具備GPS定位系統的車子逐漸普及，我們可以利用裝有GPS收發器的車輛作為探偵車取得即時交通資訊。我們的方法主要目標是估計紅綠燈時相。我們的系統TIC可以整合這些GPS資訊以及歷史統計資料，進而估算出平均速度、旅行時間、紅綠燈時相。本篇論文提出一個N-type軌跡模型，用來模擬探偵車在路上的行為估計停的時間T_red、通過的時間T_green、停的位置D_stop，可以利用N-type軌跡模型算出停的時間、通過的時間以及停的位置。我們提出一個演算法根據T_red、T_green的數量求出紅綠燈的週期，再根據T_red、T_green、D_stop的關係，提出另一個演算法去估計綠燈起始時間和紅燈起始時間。我們以光復路為實驗路段，將紅綠燈時相真值和估計值去做誤差統計，實驗結果顯示在沒有賽車的軌跡會有準確的紅綠燈時相估計，最後我們的系統可以估計紅綠燈時相和修正紅綠燈位置將拓展至二維路段。

With the substantial increase of users use urban roads, traffic jams lead to increasing travel time to road users. Traffic signals can be a great impact on travel time. Since cars with GPS positioning systems are becoming common, we can obtain traffic information using probe cars that are equipped with GPS receivers and wireless commuinicaiton capacities. An aim of this thesis is to determine traffic signal phase. Our system analyzes probe cars' historic GPS data to estimate the traffic signal phase. This thesis uses N-type trajectory model to extract the stop-and-go behaviors of probe vehicles corssing signaled intersections. The N-type trajectory model was used to estimate the stop time, pass time and stop position when a probe car stopped for a red light before crossing. We developed an algorithm that can compute traffic cycle time based on the accumulated counts of stoppings and passings of probe cars in each 5-sec. time slot of a day. We also developed an algorithm to estimate the start time of green light and start time of red light based on the stop position, stop time and pass time. We chossed GuangFu Road as our experiment road and show the errors between true values and estimated values of the traffic light phases. The results of our experiments indicate accurate estimation of traffic signal phase can be obtained with a small number of stop-and-go trajectories at the target intersection. Our estimations of the traffic signal phase are more accurate for green light phase.