基于轨迹数据的车辆跟驰行为分析与建模综述

Review of Car-following Behavior Analysis and Modeling Based on Trajectory Data

随着轨迹收集技术与数据分析技术的迅速发展,越来越多的车辆行驶轨迹被采集并用于交通流研究。车辆轨迹数据主要包括车辆运行的位置与时间等信息,利用这些信息可以推算出车辆的速度、加速度及其与前车之间的空间和时间距离等驾驶行为参量。通过研究轨迹数据可以揭示车辆自身的运行规律,车辆之间的相互作用规律,道路环境对车辆的作用规律,以及由此产生的宏观、微观交通流现象,因此,轨迹数据研究受到日益重视。本文简要回顾了与轨迹数据收集相关的历史,介绍了自然场景下采集的Next Generation SIMulation(NGSIM)数据及实验场景下采集的车队轨迹数据,并梳理了近几年基于车辆跟驰轨迹的理论研究。首先,分析以交通振荡、交通回滞为代表的交通流关键实测现象研究工作;整理跟驰行为分析方面的研究成果,包括不对称跟驰行为、稳定跟驰行为的存在性、跟驰行为的记忆效应、任务难度、随机性、异质性。之后,介绍基于跟驰行为分析成果而构建的仿真模型。最后,从3个方面评述现有基于轨迹数据的研究,并提出未来展望:交通流关键实测现象方面,应收集更多不同条件下的数据,并尝试构建更加普适性的理论或模型解释交通流现象;跟驰行为分析方面,可结合数据挖掘技术或生理、心理理论,量化驾驶员跟驰特性与生理、心理特征,并将两者结合深入分析跟驰行为的机理;仿真建模方面,可更多考虑驾驶员生理和心理变量,使模型更具人性化特征,并关注模型的评价方法,注重模型对实际交通流的解释能力。

With the rapid development of trajectory collection technology and data analysis technology,more and more vehicle trajectories are collected and analyzed in traffic flow research.Vehicle trajectory data mainly includes the location and time of the vehicles,which can be used to calculate the driving behavior parameters,such as the speed,acceleration and space headway or time headway between the vehicle and vehicle in front.Through the analysis of trajectory data,the operation law of vehicles themselves,the interaction law between vehicles,the law of road environment on vehicles and the resulting macro and micro traffic flow phenomenon can be revealed,so the research of trajectory data has been paid more and more attention.The paper briefly reviews the history of trajectory data collection,introduces Next Generation SIMulation(NGSIM)data collected in natural scenes and Vehicle Platoon Data collected in experimental scenes,and combs the micro theoretical research based on the car-following trajectory in recent years.Firstly,the research work of key phenomena in traffic flow,such as traffic oscillation and traffic hysteresis,is analyzed;The research results of car-following behavior analysis are summarized,including asymmetric car-following behavior,the existence of stable car-following behavior,the memory effect,task difficulty,stochasticity and heterogeneity of car-following behavior;Then,the introduction of simulation models based on the above analysis results of car-following behavior is followed.Finally,from these three aspects,some discussions and prospects are made according to the research status:In the aspect of key phenomena of traffic flow,more data under different conditions should be collected,and more universal theories or models should be built to explain traffic flow phenomena;In the aspect of car-following behavior analysis,data mining technology or physiological and psychological theories can be used to quantify the car-following characteristics and physiological or psychological characteristics,and combine them to deeply analyze the mechanism of car-following behavior;In the aspect of simulation modeling,the physiological and psychological variables of drivers can be considered more in the future to make the model more humanized;At the same time,pay attention to the evaluation method of the model and the interpretation ability of the model to the empirical traffic flow.