王文璇
恶劣天气会对快速道路交通安全造成显著的影响，对于传统人工驾驶车辆而言，恶劣天气由于其突发性、局部性和常见性会极大地影响驾驶人的心理和生理表现、车辆以及道路的性能，从而对交通流的安全产生影响。
随着人工智能、移动互联等信息技术的应用，车辆的巡航系统和驾驶辅助操作系统等逐渐被应用，可以实现车车通信和车路通信的智能网联车辆已逐步出现在道路上，智能网联车辆在恶劣天气下的安全驾驶成为一个值得关注的问题。
论文采用国内外相关数据，通过数据分析以及仿真模拟的方法，以车辆微观行为作为切入点，构建了智能网联车辆跟驰模型，分析了智能网联车辆对交通流中微观的车辆跟驰行为和宏观的交通流基本图带来的影响，提出适用于智能网联车辆风险感知的安全替代指标，并针对大雨天气快速道路车辆可变限速控制策略的关键问题进行探索和研究，为提升恶劣天气下交通安全水平提供技术支撑，为不久的将来智能网联车辆交通流的安全评估以及车辆管理控制提供理论基础。从工程应用的角度来看，论文研究有助于提升未来智能网联环境下快速道路的交通安全管理水平并减少恶劣天气下快速道路上存在的安全隐患，具有重要的工程应用价值。
论文的主要研究内容和成果如下：
首先，利用挪威和中国采集的恶劣天气下车辆跟驰行为数据分析恶劣天气对驾驶行为的影响。基于挪威采集的交通流数据和天气数据对比分析晴天、雨雪天和不同地面状态对跟驰车速、跟车间距和跟车时距等指标的影响；然后基于上海自然驾驶数据对比分析晴天、雨天、大雾天对跟驰车速、跟车间距和跟车时距的影响。对比两个数据集的结果得到普适性结论。
其次，根据智能网联车辆与传统人工驾驶车辆跟驰行为的不同机理，构建符合其行为规律的跟驰模型。在“车车通信”的环境下，智能网联车辆行为会受到前方多辆车的影响，以传统人工驾驶车辆跟驰模型为基础，考虑多前车与本车的速度差、前车加速度以及不同的反应时间构建智能网联车辆跟驰模型。根据车辆轨迹数据对人工驾驶车辆跟驰模型参数进行标定；对跟驰模型的稳定性进行分析；根据车队跟驰轨迹从安全、效率、能耗三个角度评价智能网联车辆对交通流的影响，提出获取模型参数最优取值的框架和确定最大车队规模的方法；对混合交通流进行仿真，分析通信车辆数量以及车队数量对交通安全的影响。
然后，以跟驰模型为基础，进一步探究智能网联车辆对于人工驾驶车辆和智能网联车辆组成的混合交通流宏观基本图，即流量-密度-速度关系图的影响。其中，考虑了智能网联车辆跟随人工驾驶车辆时出现的无法发挥功能而退化的情况。对交通流基本图进行确定性建模，分析智能网联车辆渗透率和组队强度对基本图的影响；对交通流基本参数进行敏感性分析。考虑不同类型车辆驾驶人反应时间的异质性，对交通流基本图进行不确定性建模，通过仿真的方式对获取稳定基本图需要的仿真车辆数量进行分析，同时探究渗透率和组队强度对基本图离散度的影响。
其后，基于车辆轨迹数据，对不同车型组合跟驰行为的安全替代指标分布进行比较，以跟车间距和跟车时距为例，探究安全替代指标随车速的变化规律，并寻找指标在不同速度区间内的最优拟合分布，从而得到前后车辆车型对跟驰行为的影响机理。根据风险稳态理论，筛选稳定跟驰片段，探究不同车型组合的跟驰行为安全替代指标异质性以及驾驶人群体激进度异质性；进一步地，比较不同安全替代指标，筛选出可以用于智能网联车辆风险预警的安全替代指标并确定其安全阈值。基于风险稳态理论和风险场论，提出适用于风险感知的安全替代指标，使智能网联车辆具有与驾驶人风格相似的风险感知能力，新指标具有更高的风险预测准确率、时效性、鲁棒性等优势。
最后，针对大雨天气带来的交通安全隐患，考虑受影响的能见度和地面摩擦系数，分别对人工驾驶车辆和智能网联车辆设计出不同的可变限速控制方法，从而降低交通流的风险水平。利用大雨天气下的交通流数据标定雨天的车辆跟驰模型，搭建符合雨天交通行为的交通仿真平台。根据受到雨天影响的驾驶人视距和车辆跟车间距的关系以及车辆制动性能，将不同激进度的驾驶风格引入可变限速的计算方法，得到不同驾驶风格对应的合理限速值。对雨天不同驾驶风格的人工驾驶车辆和智能网联车辆分别采取不同限速策略后在交通安全和交通效率方面的控制效果进行比较。
关键词：跟驰模型，交通流基本图，混合交通流，安全评价，可变限速控制
Adverse weather will have a significant impact on expressway traffic safety. For traditional human-driving vehicles, adverse weather will greatly affect the driver's psychological and physiological performance, vehicle performance and road performance due to its suddenness, locality and commonness, thus affecting the safety level of traffic flow.
With the application of new generation information technology such as artificial intelligence and mobile internet, vehicle cruise system and advanced driving assistance system are gradually applied. Connected and autonomous vehicles (CAVs) that can realize vehicle to vehicle (V2V) communication and vehicle to infrastructure (V2I) communication have gradually appeared on the road. The safe driving of CAVs in adverse weather deserves our attention.
The paper adopts relevant data at home and abroad, applies the data analysis and simulation methods, and takes the micro behavior of vehicles as the starting point do the research. The paper constructs the car-following model of CAVs to investigates the influence of CAVs on the micro behavior and macro fundamental diagram; proposes a safety surrogate measure which is suitable for risk perception and evaluation of CAVs; explores the key problems of variable speed limit control strategy of expressway CAVs in adverse weather. This study provides technical support for improving the level of traffic safety in adverse weather and builds a basis for traffic safety assessment and vehicle management control for the mixed traffic in the near future. From the perspective of engineering application, the research of this paper will help to improve the traffic safety management level of expressway in the future connected and autonomous environment and reduce the potential risk of expressway in adverse weather, which has an important engineering application value.
The main research contents and results of this paper are as follows:
Firstly, the data of car-following behavior collected in Norway and China in adverse weather is analyzed to explore the difference of driving behavior caused by adverse weather. According to the traffic flow and weather data collected in Norway, the changes of driving behavior such as speed, distance headway and time headway are compared and analyzed in clear/rainy/snowy weather with different road conditions. Then, the impact of clear/rainy/foggy weather from natural driving data of Shanghai on the speed, distance headway and time headway are analyzed. By comparing the results of the two datasets, universal conclusions are acquired.
Secondly, according to the different mechanism of car-following behavior between CAVs and traditional human-driving vehicles, a car-following model for CAVs in line with its behavior law is constructed. In V2V environment, the behavior of CAVs will be affected by multiple preceding vehicles, thus the car-following model of CAVs is build based on the traditional car-following model and considers the speed difference between multiple preceding vehicles and the ego vehicle, the acceleration of multiple preceding vehicles and the different reaction time. The car-following model of human driving vehicles is calibrated with the vehicle trajectories and the stability of the model is explored. The impact of CAVs on traffic flow is evaluated from the perspectives of safety, efficiency and energy consumption to acquire the framework for obtaining the optimal value of model parameters and the methof to acquire the maximum platoon size. The mixed traffic flow is simulated to analyze the impact of different number of communication vehicles and platon size.
Thirdly, based on the car-following model, the impact of CAVs on mixed macro traffic flow fundamental diagram of human-friving vehicles and CAVs, namely the relationhisp among the traffic flow, density and speed. and heterogeneous traffic flow fundamental diagram are analyzed respectively. The degradation of CAVs when following a human-driving vehicle is considered. The deterministic modeling for the traffic flow fundamental diagram is carried out to analyze the influence of CACC penetration rate and platoon intensity on the fundamental diagram. The sensitivity of basic parameters is investigated. Considering the heterogeneity of driver's reaction time, the stochastic modeling of traffic flow fundamental diagram is carried out, and the number of simulation vehicles required to obtain a stable fundamental diagram is simulated and analyzed. At the same time, the effects of penetration rate and platoon intensity on the fundamental diagram are also explored.
Fourthly, for the vehicle trajectories of different vehicle type combinations in the traffic flow, taking the distance headway and time headway as examples, explore the variation law of the indicator with the vehicle speed, find the optimal fitting distribution in different speed ranges and obtain the influence mechanism of different vehicle type combinations on the car-following behavior. Then, according to the risk homeostasis theory, the stable car-following fragments are screened, the drivers' preference for risk selection of different vehicle combinations is explored from the perspective of safety surrogate measures, and the car-following heterogeneity caused by vehicle combinations is verified from the perspective of safety surrogate measures and the index of group aggressiveness of drivers. Further, by comparing the performance of different safety surrogate measures, the threshold of a SSM in different vehicle combinations is determined, which can be applied to the early risk warning of CAVs. Based on the risk homeostasistheory and field theory, a safety surrogate measures based on risk perception is proposed. The new indicator can more accurately match the driver's perception and behavior of risk, and has higher risk prediction accuracy, timeliness, robustness and so on.
Fifth, for potential traffic risk caused by heavy rain, considering the affected visibility and ground friction coefficient, different variable speed limit control methods are designed for human-driving vehicles and CAVs respectively, so as to reduce the risk level of traffic flow. The traffic flow models in clear and adverse weather are calibrated respectively to build the simulation platform to describe the drivers’ behavior in adverse weather. According to the relationship between driver's sight distance affected by adverse weather and space distance, considering the vehicle braking performance, different driving styles are introduced into the calculation method of variable speed limit, and the reasonable speed limit values corresponding to different driving styles are obtained. The control effects of different speed limit strategies on traffic safety and traffic efficiency of human-driving vehicles and CAVs with different driving styles in adverse weather are compared.
Key Words: car-following model, fundamental diagram of traffic flow, mixed traffic, safety evaluation, variable speed limit control
