移动机器人避障的机器学习算法研究  

Research on Machine Learning Algorithm for Obstacle Avoidance of Mobile Robots

作　　者：宋启松 李少波[1] 李志昂 胡建军 SONG Qi-song;LI Shao-bo;LI Zhi-ang;HU Jian-jun(School of Mechanical Engineering,Guizhou University,Guizhou Guiyang 550025,China)

机构地区：[1]贵州大学机械工程学院,贵州贵阳550025

出　　处：《机械设计与制造》2022年第3期272-275,共4页Machinery Design & Manufacture

基　　金：国家智能制造新模式应用项目(工信厅装函[2017]468号);贵州省科技计划项目(黔科合人才[2015]4011、黔科合平台人才[2016]5103、黔科合平台人才[2017]5788)。

摘　　要：为提高移动机器人的实时避障能力,采用机器学习算法建立机器人实时判别模型,可根据检测到的障碍物距离实时预测出机器人移动方向。通过对机器学习算法的综合评价指标进行研究,分析模型预测性能的好坏。同时,为防止模型过拟合,减小泛化误差,再对机器学习算法进行交叉验证。实验结果表明,随机森林算法的综合评价指标最高,均达到了99.9%,朴素贝叶斯综合评价指标最差,约在52.9%,机器学习算法在交叉验证之后可生成更加可靠稳定的实时判别模型。In order to improve the real-time obstacle avoidance ability of mobile robots,a robot learning real-time discriminant model is established by using machine learning algorithm,the moving direction of the robot can be predicted in real time according to the distance of the detected obstacles.By studying the comprehensive evaluation index of machine learning algorithm,the prediction performance of the model is analyzed.At the same time,in order to prevent the model from overfitting and reduce the generalization error,the machine learning algorithm is cross-validated. The experimental results show that the comprehensive evaluation index of the random forest algorithm is the highest,reaching 99.9%,and the naive Bayesian comprehensive evaluation index is the worst,about 52.9%.The machine learning algorithm can generate a more reliable and stable discriminant model after cross-validation.

关 键 词：移动机器人 机器学习算法 综合评价指标 交叉验证 

分 类 号：TH16[机械工程—机械制造及自动化] TP242[自动化与计算机技术—检测技术与自动化装置]