基于CarSim的驾驶员转向特性辨识模型研究.DOCVIP

汽车驾驶员转向特性分类与辨识方法研究 李刚，韩海兰，袁航，周致成 (辽宁工业大学汽车与交通工程学院，辽宁锦州，121001) 摘要：针对汽车驾驶员转向特性分类与辨识问题，论文基于CarSim仿真平台对研究方法进行了初步探索。设计了转向工况仿真试验，采集试验数据，根据车辆最大横摆角速度使用K-means聚类算法对驾驶员转向特性进行分类；在Matlab软件环境下分别采用学习向量量化（LVQ）神经网络、BP神经网络、支持向量机（SVM）建立驾驶员转向特性辨识模型，对三种网络建立的辨识模型进行测试试验和比较。试验结果表明:三种辨识方法均具有较高的辨识精度，其中支持向量机方法在汽车驾驶员转向特性辨识方面有一定的优势。 关键字：驾驶员转向特性；辨识模型 ；CarSim ；仿真平台 中图分类号:U461.9　　 文献标识码:A　 Study on Classification and Identification Methods of Driver Steering Characteristics Li Gang,Yuan Hang,Zhou Zhi-cheng,Han Hai-lan ( College of Automobile and Traffic Engineering, Liaoning University of Technology, Jinzhou Liaoning 121001 ,China) Abstract : For the vehicle driver’s steering characteristic classification and identification problem,the research methods were initially explored based on CarSim simulation platform. The simulation experiments of steering condition were designed and the test data was collected. According to the maximum yaw rate of the vehicle, the driver steering characteristics were classified using K-means clustering algorithm. The learning vector quantization (LVQ) neural network, BP neural network, support vector machine (SVM) in the environment of Matlab software were used to establish the driver steering characteristics identification model respectively.The test experiment and comparisons were done for the three kinds of method in the paper. The results showed that three kinds of identification methods have a high accuracy and and the SVM method has a certain advantage in the aspects of driver steering characteristics identification. Keywords : Driver steering characteristics ;Identification model; CarSim ; Simulation platform 1 引 言 随着汽车保有量的不断增加，车辆的安全性、舒适性倍受人们关注。在人-车-路闭环系统中驾驶员的操纵行为占主导地位，更深入的研究驾驶员的驾驶行为有利于通过汽车电控系统提高车辆的安全性和舒适性。 参考国内外相关文献，在驾驶员行为分类及辨识方面国内外还都处于起步阶段。文献[1]建立了驾驶员实时监控系统，提出了操纵危险系数这一概念，根据ABS、ESC、TCS等汽车稳定控制系统提供的信息分析当前驾驶条件和操纵危险系数间的关系，将驾驶员分为鲁莽型、专业型、一般型和谨慎型。Wataru Takano等人根据隐形马尔科夫理论，提出了一种驾驶员转向意图辨识的智能认知方法；该方法主要包含数据分段、时间序列数据标号、辨识及产生驾驶模式[2]。Raksincharoensak P等人提出了一种基于驾驶