基于峰值检测的跟驰瞬时反应时间估计方法

Reaction time is one of the key parameters in car-following models. To address the inability of current trajectory-based reaction-time extraction methods to effectively handle randomness, a new method for instantaneous reaction-time estimation based on peak detection is proposed. First, based on the stimulus-response theory, the peak detection algorithm is used to capture the local peaks of relative speed and acceleration in the trajectory data. Then, the minimal cost time function is added to match the stimulus-response relationship, whereby the instantaneous reaction time can be estimated. The effectiveness and reliability of the method are validated using the high-precision trajectory dataset, Zen Traffic Data. The time-varying characteristics of the reaction time are further explored at the level of individual vehicles and the whole population, respectively. The experimental results show that the mean difference between the estimated results and those obtained by existing methods does not exceed 0.1 s, with both distributions being identical to actual results. The analysis of the traffic state reveals that the reaction times are concentrated in the ranges of 0.4~1.0 s under smooth condition and 0.5~1.5 s under congestion condition. While the analysis of the vehicle driving state reveals that the reaction time is relatively stable under a uniform driving state, with the average value of 0.94 s. The study demonstrates that the method not only accurately estimates the reaction time but also reveals significant differences in the reaction times under various congestion conditions and driving states, providing an effective tool for driving safety research.