Abstract: The characteristics of time headway distribution of left-turn traffic flow in peak and non-peak periods at Furong Middle Road: Renmin Middle Road — a typically signalized intersection in Changsha were analyzed under saturated and compound conditions respectively.The calculation of measured data at intersections shows that the prediction results of left-turn traffic flow saturated time headway by common methods such as CJJ37 in China and HCM2010 in America are larger than others, and the fitting results of four traditional headway distribution models and nine basic curve models in SPSS are not ideal. Piecewise function was used to establish the unified models under saturated and compound conditions respectively.By adjusting the calibration coefficient and growth coefficient of the model, the characteristics of time headway distribution in peak and non-peak periods were obtained respectively.The test analysis of the measured data shows that:over 27% of the vehicles share the highest percentage of the distribution of time headway in 2.000 to 2.500 seconds and over half of them stay in 1.500 to 2.500 seconds under saturated and compound conditions.The concentration of time headway distribution in non-peak period is lower than that in peak period, and tends to concentrate on smaller time headway.The prediction errors of the commonly used methods are usually more than 16%, while the errors of the proposed model are less than 4%.Compared with the traditional model, the proposed model can better fit the actual distribution of time headway, and the fitting effect is more ideal.

Key words: traffic engineering, distribution model, modeling analysis, time headway, left-turn flow, signalized intersection, peak hour

摘要： 针对长沙市典型信号交叉口芙蓉中路-人民中路，分别在饱和状态和复合状态下对高峰与非高峰两个时段左转车流的车头时距分布进行特性分析。交叉口实测数据验算表明，我国现行规范CJJ37和美国规范HCM2010等常用方法对左转车流饱和车头时距预测结果均偏大较多，采用4种传统的车头时距分布模型和SPSS中的9大基本曲线模型的拟合效果均不理想。采用分段函数分别建立了饱和状态和复合状态下的统一模型，通过调整模型中的标定系数和增长系数，即可分别得到高峰时段和非高峰时段的车头时距分布特征。实测数据检验分析表明：饱和状态和复合状态的车头时距分布都存在2.000~2.500 s内集中程度最高且超过27%，集中在1.500~2.500 s内超过50%的现象；非高峰时段车头时距分布集中程度低于高峰时段，且倾向于集中在较小的车头时距上；目前常用方法对车头时距平均值的预测误差通常都超过16%，文中模型误差则低于4%，与传统模型相比能较好地拟合车头时距的实际分布形态，拟合效果更为理想。

关键词: 交通工程, 分布模型, 建模分析, 车头时距, 左转车流, 信号交叉口, 高峰时段