APP下载

基于SOM-Kmeans算法的司机驾驶风格研究

2023-04-21罗雲潇张海瑞张振京宋业栋屈亚祥

时代汽车 2023年8期

罗雲潇 张海瑞 张振京 宋业栋 屈亚祥

摘 要:本文以车辆历史运行物理参数为研究对象,使用SOM-Kmeans聚类模型识别出司机的驾驶风格,为发动机经济优化提供实际指导意义。首先基于K-means聚类优先识别出了九种行驶工况,从中选取加速行为对应的三类标签以驾驶循环为单位做特征统计;随后利用因子分析对数据降维,并通过SOM-Kmeans模型进行聚类,得到温和型、普通型和激进型三种类别的驾驶风格。

关键词:行驶工况 驾驶风格 因子分析 SOM-Kmeans

Study on Driving Style of Drivers based on SOM-Kmeans Algorithm

Luo Yunxiao Zhang Hairui Zhang Zhenjin Song Yedong Qu Yaxiang

Abstract:This paper takes the physical parameters of vehicle historical operation as the research object, and uses the SOM-Kmeans clustering model to identify the driving style of drivers, providing practical guidance for engine economic optimization. Firstly, nine driving cycles were identified preferentially based on K-means clustering, and three types of tags corresponding to acceleration behaviors were selected to make feature statistics in driving cycles. Then factor analysis was used to reduce the dimension of the data, and the SOM-Kmeans model was used for clustering, and three types of driving styles, mild, ordinary and radical, were obtained.

Key words:driving cycles, driving style, factor analysis, SOM-Kmeans

1 引言

卡车、工程车等大型车的高耗油量使众多企业运营成本居高不下,且鉴于能源危机与全球变暖愈发严重,节省燃油已成为全球共识[1]。本文通过对司机在驾驶过程中的加速行为进行分析,辨识不同司机驾驶风格,针对不同驾驶风格优化油门踏板MAP,为节省油耗提供方法支持。汪益纯等[2]从交通安全出发,根据实际案例构建出影响初驾者的驾驶行为类别及其差异性;黄斐等[3]采用问卷调查的方式,按照因子分析与AHP相结合的模型对驾驶员倾向性进行建立评价体系进行辨识;吕明等[4]从对车辆的性能要求出发,使用SOM神经网络对起步工况进行聚类得到三种风格标签:温和型、普通型和激进型;王科银等[5]使用SVM驾驶风格识别模型方法与ANN模型进行了对比,得到SVM模型识别精度更高;姚柳成等[6]先使用相关分析与主成分分析对数据进行筛选与降维,再根据K-means算法对驾驶行为进行分类辨识,也得到三种驾驶行为风格。……

登录APP查看全文