Abstract: Traffic congestion not only increases the average travel time, but also increases the travel time uncertainty. It is important to study the uncertainty of travel time to calculate the congestion cost and help the traveler to make a travel plan. Based on the assumption that the travel time length distribution follows the lognormal distribution, 90% travel time length can be expressed as a function of average travel time length and coefficient of variation (COV). By exploring taxi GPS data in Shenzhen, it is found that the fluctuation of COV is less than 10%, which has little effect on 90% travel time length; and 90% travel time length shows a significant linear relationship with the average travel time. The results indicate that the travel time uncertainty linearly rises with the traffic congestion degree. The conclusion provides a theoretical and empirical support for computing the travel time uncertainty.

Key words: traffic engineering, traffic congestion, travel time uncertainty, taxi GPS data, lognormal distribution

摘要： 交通拥堵不仅延长了平均出行时长，同样增加了出行时长不确定性。研究出行时长不确定性对计算拥堵成本、帮助出行者制定出行规划具有重要意义。基于出行时长分布服从对数正态 分布假设，提出90%分位数出行时长可表示为平均出行时长和变异系数的函数。通过挖掘深圳市出租车GPS信息，发现变异系数的波动在10%以内，因此对90%分位数出行时长影响很小；而90%分 位数出行时长与平均出行时长呈现明显的线性关系。研究表明出行时长不确定性会随着拥堵程度增加而呈线性增长，该结论为计算出行时长不确定性提供理论和实证支持。

关键词: 交通工程, 交通拥堵, 出行时长不确定性, 出租车GPS数据, 对数正态分布