基于极限学习机的驾驶员制动意图识别  

Recognition of Driver’s Braking Intention Based on Extreme Learning Machine

作　　者：周恒平 牛志刚[1] Zhou Hengping;Niu Zhigang(Taiyuan University of Technology,Shanxi 030024)

机构地区：[1]太原理工大学,山西030024

出　　处：《汽车技术》2021年第11期30-34,共5页Automobile Technology

基　　金：山西省科技重大专项项目(20181102009);山西省研究生联合培养基地人才培养项目(2018JD13)。

摘　　要：为了更准确、实时地识别驾驶员制动意图,实现稳定的再生制动功能,通过分析车辆行驶状况和驾驶员操作特性,提出了一种基于邻域成分分析(NCA)和极限学习机(ELM)算法的制动意图分类与识别方法。对制动相关的特征参数进行邻域成分分析,选取制动踏板位移及其变化率和制动踏板力3个识别参数,运用极限学习机理论建立包含单隐层神经网络的制动意图识别模型,对其进行优化并利用试验数据开展验证。结果表明,模型识别准确率达到95.56%,用时为0.2 s,提出的制动意图识别方法具有较好的识别准确率和实时性。In order to identify the driver’s braking intention more accurately and in real time then achieve stable regenerative braking function,this paper,by analyzing the vehicle driving conditions and driver’s operating characteristics,proposes an braking intention classification and recognition method based on Neighborhood Component Analysis(NCA)and Extreme Learning Machine(ELM)algorithm.Firstly,the NCA is performed on the brake-related characteristic parameters to select 3 identification parameters,including the brake pedal displacement,its change rate and brake pedal force.Then,the braking intention recognition model including a single hidden layer neural network is established by using the theory of ELM,which is optimized and verified with experimental data.The test results show that the model recognition accuracy rate reaches 95.56%and the time is 0.2 s,the proposed braking intention recognition method has better recognition accuracy and efficiency.

关 键 词：制动意图 再生制动 邻域成分分析 极限学习机 

分 类 号：U461.3[机械工程—车辆工程]