基于峰值检测的跟驰瞬时反应时间估计方法  

Estimation method of car-following instantaneous reaction times based on a peak detection algorithm

作　　者：刘婧 王扬 刘冬梅[2] 刘娉婷 贺正冰 LIU Jing;WANG Yang*;LIU Dongmei;LIU Pingting;HE Zhengbing(College of Metropolitan Transportation,Beijing University of Technology,Beijing 100124,China;Research and Development Center of Transport Industry of Big Data Processing Technologies and Application for Comprehensive Transport(HSTG),Research Institute of Highway Ministry of Transport,Beijing 100088,China;School of Traffic and Logistics,Southwest Jiaotong University,Chengdu 611756,China;Laboratory for Information and Decision Systems,Massachusetts Institute of Technology,Cambridge 2139-4307,USA)

机构地区：[1]北京工业大学,城市交通学院,北京100124 [2]交通运输部公路科学研究院,综合交通运输大数据处理及应用技术交通运输行业研发中心(中路高科),北京100088 [3]西南交通大学,交通运输与物流学院,成都611756 [4]麻省理工学院,信息与决策系统实验室,美国剑桥2139-4307

出　　处：《交通运输工程与信息学报》2024年第4期85-95,共11页Journal of Transportation Engineering and Information

基　　金：国家自然科学基金项目(10038001201901)。

摘　　要：反应时间是车辆跟驰模型的关键参数之一,为解决现有基于车辆轨迹的反应时间提取方法无法有效处理随机性等问题,提出了一种基于峰值检测的瞬时反应时间估计新方法。首先,基于刺激-反应理论利用峰值检测算法捕捉轨迹数据中相对速度和加速度的局部峰值,然后结合最小费用时间函数匹配刺激-反应关系进而估计瞬时反应时间。应用高精度轨迹数据集Zen Traffic Data验证方法的有效性和可靠性,并分别从单个车辆和总体的层面进一步挖掘驾驶员的时变反应特性。实验结果显示,新方法与现有方法估计结果的均值差异不超过0.1 s,且与实际分布一致。对交通状态分析发现,反应时间在畅通条件下集中分布在0.4~1.0 s,在拥堵条件下集中分布在0.5~1.5 s;对车辆行驶状态分析发现,反应时间在匀速驾驶状态下较稳定,均值为0.94 s。研究表明,基于峰值检测的新方法不仅能够准确估计反应时间,而且能够揭示反应时间在不同交通拥堵条件和行驶状态下的显著差异,为车辆行驶安全研究提供有效工具。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.

关 键 词：交通工程 跟驰行为 瞬时反应时间 反应特性 峰值检测算法 

分 类 号：U491[交通运输工程—交通运输规划与管理]