基于特征聚类驾驶员服从度跟驰模型参数标定  

Driver Compliance Model Parameter Calibration Method Based on Feature Clustering

作　　者：陈钉均[1,2,3] 梁芮嘉 CHEN Ding-jun;LIANG Rui-jia(School of Transportation and Logistics,Southwest Jiaotong University,Chengdu Sichuan 610031,China;National and Local Joint Engineering Laboratory of Comprehensive Intelligent Transportation,Southwest Jiaotong University,Chengdu Sichuan 610031,China;National Engineering Laboratory of Integrated Transportation Big Data Application Technology,Southwest Jiaotong University,Chengdu Sichuan 610031,China)

机构地区：[1]西南交通大学交通运输与物流学院,四川成都610031 [2]综合交通运输智能化国家地方联合工程实验室,四川成都610031 [3]综合交通大数据应用技术国家工程实验室,四川成都610031

出　　处：《计算机仿真》2024年第10期126-132,共7页Computer Simulation

基　　金：国家自然科学基金项目(52072314,52172321,52102391);四川省科技计划项目(2020YJ0268,2020YJ0256,2021YFQ0001,2021YFH0175,2022YFH0016),中国国家铁路集团有限公司科技研究计划项目(P2020X016,2019F002),中国神华能源股份有限公司科技项目(CJNY-20-02),中国铁路北京局集团有限公司科技研究开发计划课题(2021BY02,2020AY02);国家重点研发计划(2017YFB1200702);项目国家自然科学基金(71971182)。

摘　　要：服从度是指网联车驾驶员在跟驰过程中对前车传递的信息服从程度。现有的服从度划分方法无法满足精确标定跟驰模型参数的要求。提出了一种基于真实轨迹数据特征聚类精确划服从度的方法。首先,提取真实轨迹数据中与服从度相关的5类特征,设计基于离群点递归筛除的Kmeans++算法进行精确特征聚类,将服从度分为三类(高、一般、低);然后将前景理论效用函引入IDM模型,构建考虑服从度的改进IDM模型;最后将聚类后的轨迹数据按照2:1比例划分参数标定与验证数据集,分别对不同服从度的改进IDM模型与常用IDM模型进行标定与验证。使用车头间距相对均方根误差(RMSPE)作为参数标定与结果验证的误差量化指标。结果表明:高服从度-IDM模型、一般服从度-IDM模型与低服从度-IDM模型在参数标定与结果验证中RMSPE均值分别为:18.2%,21.9%,24.1%和23.2%,26.8%,28.1%,均低于IDM模型两个阶段RMSPE均值:29.7%和32.2%。说明上述分类方法满足精确划分服从度与标定改进跟驰模型参数的要求。The compliance degree in this paper refers to the degree to which the driver of a connected vehicle obeys the information transmitted by the preceding vehicle during the following process.The existing compliance classification methods cannot meet the requirements for accurately calibrating the parameters of the car following model.A method based on feature clustering of real trajectory data has been proposed to accurately measure the degree of conformity.Therefor,we propose a method to accurately delineate the degree of compliance based on the clustering of real trajectory data features,in this paper.First,the five categories of features related to compliance in the real trajectory data are extracted,and the Kmeans++algorithm based on recursive sieving of outliers is designed for accurate feature clustering to classify compliance into three categories(High,Normal,Low).Then,the foreground theoretical utility function is introduced into the IDM model to construct an improved IDM model considering compliance.Finally,the clustered trajectory data are divided into parameter calibration and validation datasets in the ratio of 2:1 to calibrate and validate the improved IDM model and the common IDM model with different compliance degrees respectively,and the relative root mean square error(RMSPE)of headway is used as the error quantification index for parameter calibration and result validation.The results show that the mean values of RMSPE in parameter calibration and result validation for the High compliance-IDM model,Normal compliance-IDM model,and Low compliance-IDM model are 18.2%,21.9%,24.1%,and 23.2%,26.8%,28.1%,respectively,which are lower than the mean values of RMSPE in two stages of IDM model:29.7%and 32.2%.It indicates that the classification method in this paper satisfies the requirements of accurately dividing the compliance degree and calibrating the parameters of the improved car-following model.

关 键 词：交通工程 特征聚类 信息服从度 递归筛除 精确划分 

分 类 号：U292.3[交通运输工程—交通运输规划与管理]