In order to investigate the mechanism of risk persistence effects between overtaking vehicles and different interacting vehicles, two UAVs (Unmanned Aerial Vehicles) were used to collect video data of overtaking behaviors on a mountainous highway. The two-dimensional time to collision (TTC) calculation method was improved to assess overtaking risk. And an accelerated failure time (AFT) model was developed based on the survival analysis theory, considering the differences in overtaking styles. The findings indicate that the average risk exposure duration was considerably extended (5.19 s) when the overtaking vehicle interacted with the overtaken vehicles, in contrast to interactions with oncoming vehicles in the opposite direction. Speeding driving style increased the risk exposure duration by 15% compared to conservative driving style. When the overtaken vehicle was a truck, the risk exposure duration increased by 19% and 42% compared to motorcycles and minibuses, respectively. In addition to the overtaking vehicle's speed, the lateral position difference between the two vehicles and the oncoming vehicle's speed had the greatest average effect on the risk exposure duration for the different overtaking scenarios.
目前,国内外学者大多通过模拟驾驶实验、高清摄像机拍摄等方式采集参与借道超车行为的车辆轨迹数据,利用碰撞时间(Time to collision,TTC)[3]、避免碰撞减速率(Deceleration rate to avoid crash,DRAC)[4]、车头时距(Time headway,TH)[5]和超车视距(Passing sight distance,PSD)[6]等单一或组合指标评估借道超车风险,并基于统计分析和机器学习等方法研究车辆运行状态、交通流等因素对双车道公路借道超车行为及风险的影响。Figueira等[7]通过驾驶模拟实验分析发现,被超车车辆的速度和车辆类型是影响乡村双车道公路借道超车行为的主要因素。Pawar等[8]基于广义线性模型(Generalized linear mixed model,GLMM)探究了不同时间压力驾驶条件下,跨越车道线时间和速度变异系数对借道超车碰撞概率的影响。Branzi等[3]研究发现,对向车道交通流量增大会导致双车道公路借道超车风险增加。
由于借道超车过程同时涉及车辆横向与纵向的复合运动,经典TTC指标难以准确量化其风险,故有学者提出将经典TTC计算方法拓展至二维空间。然而,现有二维TTC(Two-dimensional time to collision,TTC2D)计算方法忽略了车辆偏航角的变化,其假设两车间的相对碰撞角度相同。尽管该方法在跟驰场景中具有良好适用性,但却难以满足冲突类型多样、冲突角度变化较大的借道超车实时风险量化需求。因此,本文引入车辆之间最短碰撞距离和接近率的概念[13]对二维TTC计算方法进行改进。考虑到碰撞点处于车辆的外缘,车辆之间最短空间距离的位置坐标难以直接获取,首先,基于车辆质心距离、车辆尺寸和车辆偏航角推导最短碰撞距离;其次,结合车辆碰撞角度对车辆矢量速度进行求和,计算得到改进的二维TTC(Improved two-dimensional time to collision,ITTC2D)指标。
为验证ITTC2D指标量化借道超车风险的有效性,在提取出的数据样本中随机挑选两组车辆,将其与TTC2D、TTC x 和TTC y 等当前常用的二维TTC指标进行对比,结果如图5所示。
从图5可以看出,在两种场景下,TTC2D数值均明显高于ITTC2D,原因在于采用TTC2D计算方法时未考虑车辆自身的车体旋转,认为两车相对碰撞角度相同。然而,车辆旋转会显著改变碰撞角度,增加忽略冲突风险的可能性[14]。采用TTC x 和TTC y 组合计算方法可以同时量化两车在横、纵方向上的风险,但受限于一维TTC中后方车辆速度应高于前方车辆的应用条件,当后方车辆速度小于前方车辆时便无法捕捉两车之间的冲突风险,即图5中或时,相应的风险指标TTC x 和TTC y 将面临失效。
当前用于评价驾驶人驾驶风格的特征参数没有公认的标准,Zang等[15]的研究表明,不同交通场景下驾驶人的驾驶风格差异显著,需结合特定场景驾驶行为特征客观分析驾驶风格。本文基于文献[16,17],结合超车车辆的运动状态特征,选择占道行驶距离比例(Travel distance occupancy ratio,TOR)、横向偏移变异系数(Lateral migration coefficient of variation,LCV)和超速驾驶时间比例(Overspeed time ratio,OTR)作为山区公路驾驶人借道超车风格特征参数。TOR是超车车辆占用对向车道行驶距离与超车总距离之比,表征驾驶人对占用对向车道行驶距离的偏好性;LCV是超车车辆横向位置偏移的标准差与其均值之比,表征超车驾驶人横向驾驶行为的稳定性;OTR是超车车辆超速行驶的时间与超车总时间之比,表征驾驶人对超速超车的倾向性。
(2)基于K-means的驾驶风格分类
采用K-means算法划分不同驾驶人借道超车风格,以TOR、LCV和OTR为聚类指标,采用肘部法则确定最佳聚类数量,并将聚类结果作为风险暴露时间模型的协变量。由图7可知,当聚类数量由2增至4时,误差平方和(Sum of squared errors,SSE)值迅速下降。当聚类数量增加超过4后,SSE值降幅明显减缓,表明增加更多类别对模型分区精度的提升效果并不显著。据此,确定最佳聚类数量为4。
本文引入AFT模型检验协变量对生存时间的影响方向和程度。采用AFT模型分别对超车车辆与被超车车辆交互、超车车辆与对向来车交互两种场景下的风险暴露时间建模。根据生存时间和协变量的值,构建Weibull、Log-Logistic和Log-Normal三种常见分布形式下的生存函数,并使用赤池信息准则(Akaike information criterion,AIC)和贝叶斯信息准则(Bayesian information criterion,BIC)评估不同分布形式的参数生存模型性能。
MwesigeG, FarahH, KoutsopoulosH N. Risk appraisal of passing zones on two-lane rural highways and policy applications[J]. Accident Analysis & Prevention, 2016, 90: 1-12.
[2]
ChoudhariT, BudhkarA, MajiA. Modeling overtaking distance and time along two-lane undivided rural highways in mixed traffic condition[J]. Transportation Letters, 2022, 14(2): 75-83.
[3]
BranziV, MeocciM, DomenichiniL, et al. A combined simulation approach to evaluate overtaking behaviour on two-lane two-way rural roads[J]. Journal of Advanced Transportation, 2021(2): 1-18.
Liu Zhao-xia, Fu Rui, Niu Shi-feng,Risk assessment in overtaking scenarios using extreme value theory and intelligent and connected information[J]. Journal of Jilin University(Engineering and Technology Edition), 2025, 55(3): 925‑937.
[6]
MahmudS M S, FerreiraL, Hoque MdS, et al. Overtaking risk modeling in two-lane two-way highway with heterogeneous traffic environment of a low-income country using naturalistic driving dataset[J]. Journal of Safety Research, 2022, 80: 380-390.
[7]
LiD, PanH. Two-lane two-way overtaking decision model with driving style awareness based on a game-theoretic framework[J]. Transportmetrica A: Transport Science, 2023, 19(3): No.2076755.
[8]
FigueiraA C, LaroccaA P C. Analysis of the factors influencing overtaking in two-lane highways: a driving simulator study[J]. Transportation Research Part F: Traffic Psychology and Behaviour, 2020, 69: 38-48.
[9]
PawarN M, VelagaN R. Investigating the influence of time pressure on overtaking maneuvers and crash risk[J]. Transportation Research Part F: Traffic Psychology and Behaviour, 2021, 82: 268-284.
[10]
KarimiA, BoroujerdianA M, CataniL, et al. Who overtakes more? Explanatory analysis of the characteristics of drivers from low/middle and high-income countries on passing frequency[J]. Transportation Research Part F: Traffic Psychology and Behaviour, 2021, 76: 167-177.
[11]
PapakostopoulosV, NathanaelD, PortouliE, et al. The effects of changes in the traffic scene during overtaking[J]. Accident Analysis & Prevention, 2015, 79: 126-132.
[12]
SongY. Integrated overtaking model and safety analysis for truck platooning requirements on two-lane undivided highways[J]. Transportation Research Record, 2024, 2678(8): 1-18.
JiXiao-feng, DaiBing-you, PuYong-ming, et al. Overtaking duration model of two-lane mountainous highways based on survival analysis[J]. Journal of Transportation Systems Engineering and Information Technology, 2022, 22(6): 183-190.
[15]
VenthuruthiyilS P, ChunchuM. Anticipated collision time (ACT): a two-dimensional surrogate safety indicator for trajectory-based proactive safety assessment[J]. Transportation Research Part C: Emerging Technologies, 2022, 139: No.103655.
[16]
ChengH, JiangY, ZhangH, et al. Emergency index (EI): a two-dimensional surrogate safety measure considering vehicles' interaction depth[J]. Transportation Research Part C: Emerging Technologies, 2025, 171: No.104981.
[17]
ZangY, WenL, CaiP, et al. How drivers perform under different scenarios: ability-related driving style extraction for large-scale dataset[J]. Accident Analysis & Prevention, 2024, 196: No.107445.
[18]
EspinozaJ, DelpianoR. Statistical models of interactions between vehicles during overtaking maneuvers[J]. Transportation Research Record: Journal of the Transportation Research Board, 2023,678(4):323-333.
[19]
SabekB, SrourF J, El MendelekM, et al. Are you in the mood to pass? A study on the interplay of psychological traits and traffic on young drivers' overtaking behavior on two-lane, two-way highways[J]. Transportation Research Part F: Traffic Psychology and Behaviour, 2024, 101: 59-77.
[20]
YeS, ChenT, Oviedo-TrespalaciosO, et al. Investigating work-related distraction's impact on male taxi driver safety: a hazard-based duration model[J]. Analytic Methods in Accident Research, 2024, 44: No.100350.
JiXiao-feng, LiJin, PuYong-ming, et al. Survival analysis of car-following conflict exposure time on through-village highway[J]. Journal of Transportation Systems Engineering and Information Technology, 2025, 25(2): 349‑360.