基于极限学习机与振动响应的路面辨识技术研究

Investigation of Road Identification Based on Extreme Learning Machine and Vibration Response

提出了一种基于极限学习机与振动响应的行驶路面辨识系统,采集车辆在公路、碎石、越野路面的悬架动挠度、俯仰角速度及垂向振动加速度等振动状态数据,作为机器学习分类辨识的训练集和测试集;分析了极限学习机(Extreme Learning Machine,ELM)和支持向量机(Support Vector Machine,SVM)在用于路面辨识时的训练效率和辨识精度差异;探索了振动数据预处理方法及数据提取特征对辨识精度的影响规律.实车试验数据的辨识测试表明,基于极限学习机的路面辨识架构,能够在保证辨识精度的同时大幅降低计算量,为路面辨识算法的车载控制器实时运行奠定了重要基础.

This paper proposed a driving road condition identification system based on vehicle vibration data.Vibration data such as suspension deflection,pitch rate and vertical acceleration on highway,gravel,and off-road roads were collected as the training set and test set for machine learning classification and identification.The difference in training efficiency and identification accuracy of extreme learning machine(ELM)and support vector machine(SVM)when used in road identification was analyzed.The influence of vibration data pre-processing methods and data extraction features on identification accuracy were explored.The identification test based on actual vehicle test data shows that the road identification architecture based on the extreme learning machine can greatly reduce the computation and guarantee the identification accuracy,which lays an important foundation for the real-time operation of the on-board controller of the road identification algorithm.