基于驾驶人距离感知不确定性的跟驰行为建模

Car-following Behavior Modeling Based on Uncertainty of Driver Distance Perception

摘要:

以纵向控制模型（LCM）为基准跟驰模型，进行基于驾驶人距离感知不确定性的跟驰行为建模。假设在任意时刻驾驶人的感知车间距离误差分别服从均匀分布与截断正态分布，建立概率密度函数的参数（均匀分布的边界值、正态分布的均值与标准差）与实际车间距离、驾驶人激进性特征的函数关系，分别得到基于均匀分布的扩展模型与基于截断正态分布的扩展模型。最后，运用上海市自然驾驶数据对扩展模型进行标定。结果表明：基于均匀分布和截断正态分布的扩展模型的标定误差与验证误差均小于LCM，而且扩展模型的多次模拟仿真误差波动很小，即都可以用来描述驾驶人距离感知的不确定性；基于截断正态分布的扩展模型优于基于均匀分布的扩展模型。

Abstract:

By taking the longitudinal control model （LCM） as the basic car-following model，the car-following behavior is modeled based on the uncertainty of driver distance perception. It is assumed that the error of perceptual spacing follows the uniform distribution and the truncated normal distribution. Thereafter， the relationships between the probability density function parameters （i.e. the boundary values of the uniform distribution， the mean value and standard deviation of the normal distribution） and the factors （i.e. observed spacing values and driver’s aggressiveness characteristics） are established. Then，the extended model based on the uniform distribution and the extended model based on the truncated normal distribution are obtained.Finally， the Shanghai naturalistic driving data are employed to calibrate the extended models. It is shown that the extended models exhibit better performance than LCM， indicating that the extended models could capture the uncertainty of driver distance perception. Besides，the extended model based on the truncated normal distribution has better performance than the extended model based on the uniform distribution.