山区高速公路隧道群行车风险时空耦合机理解析及二次事故主动预控策略研究

Tunnel groups are the accident-prone segments of the freeway. As the proportion of tunnels increases on the mountainous freeways that are recently constructed or planned to construct, the traffic safety of tunnel groups catches more attention of freeway administrators. In tunnel groups, drivers alternately experience the light adaption and acceleration when exiting the tunnels, and the dark adaption and deceleration when entering the tunnels. The traffic operational conditions for the tunnel where the vehicle is traveling currently and the tunnels where this vehicle has traveled through have coupled impact on the driving behavior and crash risk within the current tunnel. Hence, there is a need to propose surrogate safety measures reflecting the spatiotemporal coupling of driving risk in the tunnel groups. The freeway safety management systems for tunnel groups in China are reactive to the crash occurrences at present, in the absence of reliable subsystems for proactive crash prevention (or proactive safety management, interchangeably). In addition, the freeway administrators are inclined to adopt the proactive crash prevention to the secondary crashes, instead of primary crashes.. Firstly, the sample data for the real vehicle test are collected and the mechanism for traffic operational conditions incurring driving risk in tunnel groups is analyzed via quasi-induced exposure. Then the relationship between traffic operational conditions and spatiotemporal coupled surrogate safety measures is developed based on the deep learning approach. In addition, the model for the evolution of traffic flow dynamics after the occurrence of a primary crash is formulated via the vehicle trajectory data from the surveillance subsystem of freeway safety management systems for tunnel group. Associated the model with the spatiotemporal coupled surrogate safety measures, a dynamic secondary-crash risk prediction model is proposed. Furthermore, based on the secondary-crash risk prediction model under a proactive crash prevention strategy, the strategy is optimized with the objectives of surrogate safety measure values and an algorithm of derivative-free methods.

隧道群是高速公路事故多发路段。随着大量新建和规划的山区高速公路隧道比增大，隧道群的行车安全问题越发引起管理者的重视。在隧道群中，驾驶人交替经历出隧道明适应、加速和进隧道暗适应、减速。车辆行驶所在隧道和已经过的前序隧道内的交通运行环境将耦合影响当前隧道内的驾驶行为和行车风险。因此需要提出反映行车风险时空耦合性的安全指标。国内目前隧道管理平台对事故识别处于被动状态，缺乏可靠的主动安全预控系统。并且较于首次事故，管理者更倾向将主动预控应用在防止二次事故中。本项目将采集行车样本，采用准相对风险暴露量方法解析隧道群交通运行环境对行车风险的作用机理。基于深度学习，建立隧道群交通运行环境与时空耦合行车安全指标的关系模型。其次，通过监控系统采集行车轨迹，建立首次事故发生后交通流动态推演模型，结合行车安全指标建立二次事故风险动态预测评估模型。最后，以二次事故风险指标为目标，通过无梯度方法，优化主动预控策略。

项目面向降低山区高速公路隧道群二次事故风险，综合环境光照状态及变化等因素，解析高速公路隧道群行车风险机理。通过建立光照状态与期望速度及感知间距的关系，提出高速公路隧道微观交通流模型。基于交通冲突分析法，建立高速公路隧道群二次事故风险动态预测模型，并进行高速公路隧道群二次事故主动预控策略评估。.项目成果提出了低交通流检测密度下的隧道群二次事故风险预测方法，并拓展了环境光照状态和微观交通流理论之间的量化关系。具体包括：1）提出了考虑不同位置光照状况差异的IDM车辆跟驰模型IDM-LLDS，模型将车辆期望速度的变化和隧道出入口亮度变化产生的明暗适应以及中间段的低亮度状态进行了关联，从而能描述光照状况对驾驶人跟驰行为随交通流的传播；2）提出了高速公路隧道群二次事故风险的预测模型，描述了由首次事故发生和隧道群沿线光照变化所形成的交通流扰动。采用交通冲突方法，即先基于考虑隧道群光照状况和相邻车道屏障效应的微观交通模型LMTID预测车辆轨迹，然后用替代安全指标DRAC量化事故风险；3）对比自适应隧道光照控制（ATLC）和智能网联车环境下的提前速度和变道诱导（ASLG）对二次事故风险的降低效果。.研究结果表明，降低隧道群路段二次事故风险，给驾驶人提供一个良好的照明环境比智能网联环境下的提前警示更有效果。将两个手段相结合更具前景，因为ASLG能及时提醒驾驶人首次事故产生的交通流波动，ATLC能平缓光照过渡和降低屏障效应。