高速公路异构交通流HDV建模及其特征

Modeling of human-driven vehicles and characteristics of heterogeneous traffic flow for freeway

为了更好地模拟高速公路中网联自动驾驶车辆(connected and autonomous vehicles, CAV)与人工驾驶车辆(human-driven vehicles, HDV)组成的异构交通流,开展高速公路异构交通流跟驰与换道行为研究。首先,以NGSIM轨迹数据集为基础分析HDV的微观交通特性,根据驾驶人在相同车头时距条件下的加、减速策略,将驾驶人分为保守型、普通型和激进型3类。其次,从驾驶人感知判断结果存在不确定性的角度,结合车辆跟驰数据分析不同类型驾驶人车头时距和速度判断误差的特性,同时引入信息效用理论模拟驾驶人对CAV认知程度的变化及其对驾驶决策的影响,在智能驾驶人模型(intelligent driver model, IDM)和对称双车道元胞自动机换道模型的基础上,提出一种改进的HDV跟驰与换道模型。最后,将改进模型与IDM模型预测结果进行对比分析,并通过MATLAB对高速公路异构交通流特征进行仿真分析。研究结果表明:改进模型能够更准确地模拟车辆跟驰行为,相比于IDM模型,保守型HDV跟驰模型平均绝对误差MAE降低24.7%,均方误差MSE降低11.9%,皮尔逊相关系数PCCs提高2.6%;普通型HDV跟驰模型MAE降低45.6%,MSE降低38.6%,PCCs提高4.0%;激进型HDV跟驰模型MAE降低41.2%,MSE降低45.9%,PCCs提高0.4%;在接近自由流状态下,HDV在车流中的占比对异构交通流的速度、流量和稳定性等特征影响较小,随着密度的增大HDV占比对交通流的影响也随之增大,直到达到临界密度,车辆组成对交通流的影响开始减小;在相同交通流密度下,HDV在车辆组成中的占比与交通流的速度、流量和稳定性呈负相关性。模型丰富了对异构交通流HDV跟驰与换道行为的研究,在高速公路HDV和CAV混行的异构交通流的交通管理和基础设施设计等方面具有一定参考价值。

In order to better simulate the heterogeneous traffic flow consisting of connected and autonomous vehicles(CAVs)and human-driven vehicles(HDVs)on highways,research on the car-following and lane-changing behaviors of heterogeneous traffic flow on highways was conducted.Firstly,based on the NGSIM trajectory dataset,the micro-level traffic characteristics of human-driven vehicles were analyzed.According to the acceleration and deceleration strategies of drivers under the same headway conditions,they were classified into three types,conservative,moderate,and aggressive drivers.Next,considering the uncertainty in drivers’perception and judgment,along with the analysis of the characteristics of the judgment errors in headway and speed by drivers with different driving styles based on vehicle following data,and simultaneously incorporating the information utility theory to simulate the changes in drivers’perception of CAVs and their impact on driving decisions,an improved HDV following and lanechanging model was proposed on the basis of the intelligent driver model(IDM)and STCA lanechanging model.Finally,the improved model was compared and analyzed against the predictive results of the IDM model,and simulation analyses of the characteristics of heterogeneous traffic flow on highways were conducted using MATLAB.The results show that significant improvements in accuracy are observed with the improved model,a reduction in the mean absolute error MAEby 24.7%,in the mean squared error MSEby 11.9%,and an increase in the Pearson correlation coefficient PCCsby 2.6%for the conservative HDV following model.For the normal type,decreases in MAEby 45.6%and in MSEby 38.6%,with an increase in PCCsby 4.0%,are documented.The aggressive type exhibits decreases in MAEby 41.2%,in MSEby 45.9%,and a marginal increase in PCCsby 0.4%,indicating a more accurate simulation of vehicle following behavior by the improved HDV model.The proportion of HDVs in traffic has a minimal impact on speed,flow,and stability under near-free flow conditions.However,as density increases,the impact of HDV proportion also grows,reaching apeak at a critical density,after which the influence of vehicle composition on traffic flow decreases.At the same traffic flow density,a negative correlation is found between the proportion of HDVs and the speed,flow,and stability of the traffic stream.The research on heterogeneous traffic flow,particularly focusing on HDV car-following and lane-changing behaviors is enriched by the model.This holds significant reference value for traffic management and infrastructure design in scenarios involving the coexistence of HDVs and CAVs on freeways.3tabs,14figs,17refs.