基于分层COX 模型的跟驰反应延迟时间生存分析

驾驶员的反应延迟时间是驾驶员跟驰行为的重要指标之一，也是跟驰模型中的重要参数之一. 为分析延迟时间与车辆运动状态、光照条件影响因素之间的关系及延迟时间的概率分布，通过实车实验得到跟驰行为延迟时间自然驾驶数据，采用Kaplan-Meier 方法进行延迟时间单因素分析并构建延迟时间分层COX模型. 结果表明：驾驶员跟驰反应延迟时间生存函数受前车加速度，前车加速度变化状态影响显著；前车加速度与延迟时间风险函数之间的关系随时间变化，需采用分层COX模型建模；前后车相对距离每增大10 m，延迟时间风险函数取值减小6.03%；前车由变速运动变为匀速运动时，延迟时间风险函数取值减小35.39%. 研究给出延迟时间风险函数与影响因素之间的定量关系，结果可应用于跟驰模型参数优化与微观驾驶行为仿真模型.

Car-following Response Delay Time Survival Analysis Based on Stratified COX Model

Driver's response delay time is one of the key parameters in the car- following behavior analysis and microscope car-following model. Experiments were conducted to collect naturalistic car-following behavior data. The Kaplan-Meier based univariate analysis method and delay time stratified COX model are employed to examine the influences of vehicle kinetic characteristics and lighting condition on drivers' response delay time. The results indicate that driver's response delay time during car- following process is statistical related with front vehicle's acceleration rate and its changing status. Acceleration of front vehicle does not satisfy the PH (Proportional Hazard) hypothesis, which means its influences on response delay time is related with time series. The results from stratified COX models demonstrate that the delay time's hazard function value decreases by 6.03% for every 10 m increase in the spacing between front and following vehicles, When front vehicle shift from variable motion to uniform motion, the delay time's hazard function value decreases by 35.39% . The result provides quantitative relation between delay time and influencing factors, which can be used in parameter optimization of car-following model and microscope traffic simulation.