考虑多前车作用势的混行交通流车辆跟驰模型

基于自动驾驶车辆(AV)和常规人驾车辆(RV)混合行驶的情况，在全速度差(FVD)模型的基础上考虑了多前车和一辆后车的车头间距、速度、速度差、加速度差等因素，建立了适用于AV和RV 2种车辆的混行车辆跟驰模型；引入分子动力学理论定量化表达了周围车辆对主体车辆的影响程度；利用RV和AV混行场景跟车数据，以模型拟合精度最高为目标，对所有参数遍历寻优，进行标定；对比分析了混行车辆跟驰模型和FVD模型控制下交通流的稳定性，解析了车速对交通流稳定性的影响；设计了数值仿真试验，模拟了城市道路和高速公路2种常见场景，分析了混行车辆跟驰模型的拟合精度。研究结果表明:考虑周围多车信息有利于提高交通流的稳定性；车辆速度越低交通流稳定性越差；考虑多车信息的分子动力学混行车辆跟驰模型可以提前获得整个车队的运行趋势，更好地模拟AV的动力学特征；与FVD模型相比，在城市道路条件下混行车辆跟驰模型中的RV平均最大误差与平均误差分别减小了0.18 m·s-1和13.12%，拟合精度提高了4.47%；与PATH实验室的ACC模型相比，在高速公路条件下混行车辆跟驰模型中的AV平均最大误差和平均误差分别减小了7.78%和26.79%，拟合精度提高了1.21%。可见，该模型可用于混行环境下AV的跟驰控制与队列控制，以及AV和RV的跟驰仿真。 

Vehicle-following model in mixed traffic flow considering interaction potential of multiple front vehicles

The mixed traffic flow consisting of automated vehicle (AV) and regular vehicle (RV) was analyzed. Based on the full velocity difference (FVD) model, a vehicle-following model for two types of vehicles (AV and RV) in mixed traffic flow was constructed by considering the factors of the headway, velocity, velocity difference and acceleration difference of multiple front vehicles and one rear vehicle. By introducing the molecular dynamics, the model also quantitatively expressed the influence degree of a surrounding vehicle on the host vehicle. According to the data collected from the vehicle-following field test mixed with AVs and RVs, the model parameters were globally optimized to obtain the highest accuracy. The stability of traffic flow for the vehicle-following model and FVD model was compared, and the influence of velocity on the stability of traffic flow was analyzed. Numerical simulation was designed to simulate the common traffic scenarios including urban areas and expressways, and the accuracy of the proposed model was analyzed. Simulation results indicate that the stability of traffic flow improves by considering the information from surrounding multiple vehicles, and the small velocity can reduce the stability. The proposed model can respond to the behaviours of the whole platoon in advance and simulate the dynamics characteristics of AVs better. In urban areas, compared with the FVD model, the average maximum error and average error of RV for the proposed model reduce by 0.18 m · s-1 and 13.12%, respectively, and the accuracy improves by 4.47%. In expressways, compared with the adaptive cruise control (ACC) model provided by PATH Laboratory, the average maximum error and average error of AV for the proposed model reduce by 7.78% and 26.79%, respectively, and the accuracy improves by 1.12%. In addition to providing model basis for AV-following control and queue control in mixed traffic flow, the proposed model can be utilized in vehicle-following behavior simulation for AV and RV. 1 tab, 7 figs, 38 refs.