Abstract:

With the vehicle identification data of intersections in Beijing, this paper explores the travel time and its reliability of urban expressway and some main road using the hypothesis test method. This paper first puts forward a data filtering algorithm named quartile division to screen the raw data. Then, it fits the travel time by four distribution functions The proposed distribution hypothesis is tested and then determined. Moreover, the travel time and its variance is settled. Then the paper analyzes the relationship between the travel time distribution and road condition which is represented by the average speed on road. The results indicate that for heavy traffic, the travel time distribution approximately follows the Weibull distribution. The travel time distribution tends to follow the logarithmic normal distribution for smooth traffic. This paper further explores the feasibility of normal distribution data fitting and parameter estimation method instead of other three methods, which shows that the error of the travel time is within an acceptable range and the error of travel time variance is relatively higher.

Key words: traffic engineering, travel time estimation, plate number recognition, travel time reliability, distribution pattern

摘要：

利用北京市道路交叉口的车牌识别数据，应用假设检验的方法研究城市快速路及主干道的旅行时间及其可靠性的分布与估计方法.首先使用四分位数据筛选法对原始数据进行了筛选，获得分析所用数据集.其次对旅行时间的分布进行了研究，提出了与旅行时间有关的分布假设并进行了相应的检验分析，随后研究了旅行时间分布与路段道路运行状况的关系.基于对北京市的数据分析可以看出，道路拥堵时旅行时间分布多接近Weibull分布，通畅时接近对数正态分布.最后，本文进一步探讨直接利用正态分布进行数据拟合以及参数估计的可行性.案例分析显示，使用正态分布估计旅行时间平均值误差较小，但对旅行时间可靠性的估计则误差较大.

关键词: 交通工程, 旅行时间估计, 车牌识别, 旅行时间可靠性, 分布形式