摘要
为了提高履带机器人对地面分类的准确率,提出一种基于时频特征和PCA-SVM的地面分类方法。对振动信号采用时域幅值和现代功率谱分析同时进行时频特征提取,并运用主成分分析法(PCA)进行时频特征的融合和简化,然后利用LIBSVM中的一对一支持向量机(SVM)程序,实现地面识别分类。控制履带机器人以2种速度在5种不同的地面上行驶,利用其上安装的惯性导航传感器采集3个方向直线加速度和三轴的角速度信号,采用本文算法和单一特征分类算法对信号分别进行时频特征处理与地面分类试验。结果表明,本文算法在机器人速度0. 02m/s时可得到更好的分类效果。该方法可为履带机器人实现更有效的地面环境感知和自身在最佳状态下的导航控制运行提供技术支持。
In order to improve the accuracy of terrain classification of tracked robot,a terrain classification method based on the time-frequency characteristics and PCA-SVM was proposed.The time-frequency characteristics of robots vibration signal were extracted by using time amplitude domain analysis and modern power spectrum analysis respectively,then these time-frequency characteristics values were simplified and fused by using principal component analysis(PCA),finally the terrain recognition classification were realized with one-on-one support vector machine(SVM)in LIBSVM program.In the process of experiment,the tracked robot was controlled to steady travel at two different speeds on five different terrains and the vibration signal were recorded by using an inertial navigation sensor installed on it.The vibration signals included the acceleration signals in three directions and angular velocity signals of three axes.The time frequency feature processing and terrain classification tests were carried out respectively by using the proposed algorithm and the single feature classification algorithm.The method presented in this paper could get better classification effect at the speed of 0.02 m/s.And it also provided a foundation for the tracked robot to achieve more effective terrain environment perception and navigation in the best state.
作者
杜习波
朱华
DU Xibo;ZHU Hua(School of Mechatronic Engineering,China University of Mining and Technology,Xuzhou 221116,Jiangsu,China;School of Mechanical and Power Engineering,Henan Polytechnic University,Jiaozuo 454000,Henan,China)
出处
《河南理工大学学报（自然科学版）》
CAS
北大核心
2019年第2期84-90,共7页
Journal of Henan Polytechnic University（Natural Science）
基金
国家863计划项目(2012AA041504)
关键词
履带机器人
地面分类
时频特征
主成分分析法
支持向量机
tracked robot
time-frequency characteristic
principal component analysis
terrain classification of support vector machine
