基于随机森林的驾驶人驾驶习性辨识策略  

Identification Strategy of Driving Style Based on Random Forest

作　　者：朱冰[1,2] 李伟男 汪震[1] 赵健[1] 何睿[1] 韩嘉懿[1] Zhu Bing;Li Weinan;Wang Zhen;Zhao Jian;He Rui;Han Jiayi(Jilin University, State Key Laboratory of Automotive Simulation and Control, Changchun 130022;Jilin University, Key Laboratory of Bionic Engineering of Ministry of Education, Changchun 130022)

机构地区：[1]吉林大学汽车仿真与控制国家重点实验室,长春130022 [2]吉林大学工程仿生教育部重点实验室,长春130022

出　　处：《汽车工程》2019年第2期213-218,224,共7页Automotive Engineering

基　　金：国家自然科学基金(51775235;51475206);国家重点研发计划项目(2016YFB0100904);吉林省自然科学基金(20170101138JC)资助

摘　　要：深入理解驾驶人驾驶习性及其表征方法,对于实现在汽车自动驾驶、辅助驾驶等不同控制系统下的人机和谐交互具有重要意义。为此,本文中提出了一种基于随机森林模型的驾驶人驾驶习性辨识策略。搭建了驾驶人驾驶数据实车采集系统,在典型跟车驾驶工况下对驾驶人驾驶习性数据进行了实时采集;根据层次聚类理论,对驾驶人驾驶习性进行了标定;在此基础上,引入随机森林模型建立了驾驶人驾驶习性辨识策略,并进行了重要性分析、模型训练和测试分析。测试结果表明,本文提出的基于随机森林模型的驾驶人驾驶习性辨识策略能有效辨识驾驶人驾驶习性,模型整体精准度可达97.1%。Understanding and identification of driver s driving style are of great significance to the human-machine harmonious interaction under different control systems such as automatic driving and assistant driving. A driving style identification strategy based on random forest model is proposed in this paper. Firstly, the driver s driving data acquisition system is set up. Based on that, the driving data of several drivers are collected in real time under typical car-following scenarios. According to hierarchical clustering theory, the driving style are “labeled”. On this basis, a random forest model is introduced to establish driving style identification strategy, and importance analysis, model training and identification test are carried out. The test results show that the driving style identification strategy based on the random forest model can effectively identify driver s driving style and the overall accuracy of the model can reach 97.1%.

关 键 词：车辆工程 驾驶习性辨识 随机森林 层次聚类 跟车工况 

分 类 号：U471.1[机械工程—车辆工程]