Abstract: This paper investigates vehicles' lane-changing behaviorson road segment adjacent to freeway tunnels and aims to improve vehicle's driving safety. Naturalistic driving tests were carried out on the three freeway tunnels' adjacent sections in Guangdong Province. The driving trajectories of the lane-changing vehicle and the surrounding vehicles are collected. Considering the heterogeneity of different drivers' perception of lane-changing risk level, a random parameter accelerated failure time (AFT) model was developed to describe the influence of environment and vehicle movement status on lane-changing duration in the tunnel adjacent sections. The results show that the random parameter in the AFT model has better goodness of fit compared to the fixed parameter AFT model. The significant influencing factors of lane-changing duration include distance to the tunnel, the speed difference to the leading vehicle, lane change direction, and distance to the leading vehicle in the target lane. The closer the position of lane-changing vehicle is to the tunnel and the closer to the vehicle in front of the target lane, the shorter of the lane-changing duration. Compared to the case that the speed of the lane-changing vehicle is greater than the leading vehicle in the original lane, the lane-chancing duration under a non-following state would increase by 7%. When the speed of the lane-changing vehicle is lower than the leading vehicle in the original lane, the lane-changing duration would increase by 20% compared to the case that the speed of the lane-changing vehicle is greater than the leading vehicle in the original lane. The study provides theoretical basis and method guidance for the improvement of traffic safety facilities on freeway tunnel adjacent sections and the development of microscopic driving behavior models.

Key words: traffic engineering, lane-changing duration, accelerated failure time model, road segment adjacent to freeway tunnel, random parameter, naturalistic driving

摘要： 为研究高速公路隧道临近段车辆换道行为，提高隧道路段行车安全水平，在广东省的3条高速公路隧道临近路段开展自然驾驶试验，采集换道车辆的行车轨迹以及周围行车环境等数据。基于采集到的车辆换道数据，采用生存分析的全参数估计方法，考虑不同驾驶人换道风险感知水平的异质性，构建随机参数加速失效时间(AFT)模型，分析隧道临近段行车环境、车辆运行状态等潜在因素对车辆换道持续时间的影响。研究结果表明：相较于固定参数AFT模型，随机参数AFT模型具有更好的拟合优度；至隧道进口的距离、起始车道前车的车速差、换道方向和至目标车道前车距离会对高速公路隧道临近段车辆换道持续时间产生显著影响；车辆换道位置距离隧 道进口越近，至目标车道前车的距离越近，换道持续时间越短；相较于换道车辆车速大于起始车道前车的情况，换道车辆处于非跟驰状态和车速小于起始车道前车时，换道持续时间分别增加 7%和20%。研究结果可为高速公路隧道临近段交通安全设施改善和微观驾驶行为模型构建提供理论依据和方法指导。

关键词: 交通工程, 车辆换道持续时间, 加速失效时间模型, 高速公路隧道临近段, 随机参数, 自然驾驶