考虑城市道路局部交通运行环境随机变化的跟驰模型  

A car-following model incorporating random variations in local traffic operating conditions on urban roads

作　　者：陈昱光[1,2] 梁子禄 胡山 杨彬 林弘灏 CHEN Yuguang;LIANG Zilu;HU Shan;YANG Bin;LIN Honghao(Faculty of Transportation Engineering,Kunming University of Science and Technology,Kunming 650500,China;School of Transportation,Southeast University,Nanjing 210096,China)

机构地区：[1]昆明理工大学交通工程学院,昆明650500 [2]东南大学交通学院,南京210096

出　　处：《安全与环境学报》2025年第4期1391-1399,共9页Journal of Safety and Environment

基　　金：国家自然科学基金项目(52462050,42277476)。

摘　　要：针对现有模型较少考虑交通运行环境拥挤情况对车辆跟驰行为的影响以及交通运行环境在行驶过程中受到外部影响随机变化的情况,试图建立更加符合不同交通运行环境的车辆跟驰模型。为此,提取速度、速度标准差、局部空间占有率三个指标,基于模糊C均值(Fuzzy C-Means, FCM)算法对交通运行环境进行聚类分析并实现有效量化。针对交通运行环境随时间变化的情况,拟合不同运行环境下的期望速度函数,引入高斯混合隐马尔可夫模型(Hidden Markov Model with Gaussian Mixture Model, GMMHMM)实现不同交通运行环境的识别及期望速度函数的转换,进而构建一种考虑不同交通运行环境下的车辆跟驰模型。最后,通过下一代模拟(Next Generation Simulation, NGSIM)轨迹数据,利用遗传算法标定模型参数。结果表明,与经典的全速度差(Full Velocity Difference, FVD)模型相比,所提出的跟驰模型能够更好地拟合车辆跟驰数据,其平均绝对误差(Mean Absolute Error, MAE)、均方根误差(Root Mean Square Error, RMSE)分别降低了35%、39%,R2提高了238%。This study tackles the limitations of existing models that inadequately consider the effects of varying congestion levels on car-following behavior,as well as the random fluctuations in traffic conditions caused by external factors.The objective is to develop a car-following model that more accurately captures the dynamics of driving behavior under different traffic scenarios.We identify three key indicators—vehicle speed,speed standard deviation,and local space occupancy—and utilize the Fuzzy C-Means(FCM)clustering algorithm to classify traffic conditions into three distinct states:congested,moderate,and free-flow.This clustering approach facilitates an effective quantification of traffic conditions.The clustering results indicate that as congestion increases,the average speed decreases,the impact of surrounding traffic intensifies significantly,and vehicle speed becomes highly variable and unpredictable.This variability makes stable,uninterrupted driving increasingly challenging.To better capture the temporal variability in traffic conditions,we fit expected speed functions for each traffic state and introduce a Hidden Markov Model with Gaussian Mixture Model(GMMHMM).This model allows us to identify transitions between traffic states and adapt the expected speed functions accordingly,thereby enabling the development of a car-following model that accurately reflects the impact of dynamically changing traffic conditions.The resulting state transition probability matrix effectively reflects the dynamic nature of traffic states.Notably,each traffic condition demonstrates strong persistence,with self-transition probabilities exceeding 0.8,indicating a high likelihood of remaining in the same state.Transition probabilities between adjacent conditions(e.g.,from free-flow to moderate or from moderate to congested)are relatively low,while those between non-adjacent states are minimal.These patterns indicate that abrupt shifts in traffic conditions are rare,consistent with the gradual transitions typically observed i

关 键 词：安全工程 交通运行环境 跟驰模型 高斯混合隐马尔可夫模型 全速度差模型 

分 类 号：X951[环境科学与工程—安全科学]