中文摘要：

基于实际的驾驶行为特性,将驾驶员的驾驶状态分为强跟驰、弱跟驰和自由流3种状态,建立能够描述这3种状态的车头时距三元混合分布模型.利用北京快速路实测数据,通过最大期望（EM）算法标定了三元混合分布模型中的参数,对3种不同驾驶行为数据进行分析.通过研究对比各个车头时距分布模型的拟合精度,结果表明,提出的车头时距混合模型的拟合结果优于威布尔分布以及对数正态分布模型.

英文摘要：

Based on the characteristics of driver＇s psychological behavior,driving behaviors were classified into strong car-following,weak car-following and free driving state.Then a ternary mixed distribution model that can describe the headway distribution of three driving states was built.The parameters of the ternary mixture distribution model were determined by expectation maximization（EM）algorithm through a case study of Beijing expressway,and the data of three different driving behavior were deep analyzed.Results show that the fitting precision of ternary mixed distribution model is the best compared with other Weibull and Lognormal models.