基于多国实测数据下的交通流模型对比研究  

Contrastive Research of Traffic Flow Model Based on Multinational Empirical Data

作　　者：刘张琦 谢耀华 李宝路 刘志广 徐志刚[1] LIU Zhangqi;XIE Yaohua;LI Baolu;LIU Zhiguang;XU Zhigang(School of Information Engineering,Chang′an University,Xi′an 710064;China Merchants Chongqing Communications Technology Research&Design Institute Co.,Ltd.,Chongqing 400067)

机构地区：[1]长安大学信息工程学院,西安710064 [2]招商局重庆交通科学研究院有限公司,重庆400067

出　　处：《公路交通技术》2022年第2期134-138,共5页Technology of Highway and Transport

基　　金：国家重点研发计划项目(2019YFB1600101)。

摘　　要：为检验最新的S3交通流模型和多个经典交通流模型对交通状态的描述精度和对不同国家交通流数据的适应性,利用模型分别对中国西安市长安大学CHD数据集、德国HighD数据集和美国PeMS数据集进行参数拟合和特性分析,利用流量、速度和密度作为输入参数,采用最小二乘方法拟合,并利用均方误差指标(MSE)对参数拟合后的模型进行性能评价。试验结果表明,S3模型在针对中、德、美三国实测数据条件下,其MSE v值分别为69、86、61,显著低于其他4种经典交通流模型(Greenshields、Greenberg、Underwood和Newell模型),S3模型可以更好地表达自由流、饱和、拥堵交通状态下流量、速度和密度三者之间耦合关系。In order to test representation accuracy of traffic state and the adaptability of traffic flow data in different countries by the latest S3 traffic flow model and several classic traffic flow models,CHD dataset of Chang’an University in Xi’an,China,the German HighD dataset and the American PeMS dataset are fitted and analyzed respectively by these models.The traffic flow,speed and density are used as input parameters.The least squares method is used to fit.The Mean Square Error(MSE)value is used to evaluate the performance of models after parameter fitting.The experimental results show that MSE v values of the S3 model under measured data from China,German and America are 69、86、61 respectively,which are significantly lower than the other four classic traffic flow models(Greenshields,Greenberg,Underwood and Newell models).The S3 model can better express the coupling relationship among flow,speed and density under free flow,saturated and congested traffic conditions.

关 键 词：交通流模型 S3模型 HighD数据集 PeMS数据集 

分 类 号：U491.1[交通运输工程—交通运输规划与管理]