基于Levenberg-Marquardt算法的串联协作机器人精度标定研究  

Research on Accuracy Calibration of Serial Collaborative Robots Based on Levenberg-Marquardt Algorithm

作　　者：金小飞 李杨 Jin Xiaofei;Li Yang

机构地区：[1]哈工大机器人(合肥)国际创新研究院,安徽省合肥市230601 [2]安徽三联学院工学部

出　　处：《工程机械》2024年第9期54-59,I0017,I0018,共8页Construction Machinery and Equipment

基　　金：安徽三联学院自然科学重点项目(KJZD2023006)。

摘　　要：针对自主研发的Ultra-T30型串联协作机器人建立四参数MDH运动学模型,利用FARO激光跟踪仪采集标定点坐标,并构造机器人几何参数集合与末端位置误差两者之间的最小二乘目标函数,基于Levenberg-Marquardt算法的关节偏移辨识,进一步对机器人几何参数进行补偿,并通过10组验证点对标定结果校核。研究结果表明:标定后机器人Ultra-T30平均误差降低了79.67%,均方根误差较标定前降低了79.75%,这也为Ultra-T30机器人的商业应用推广奠定了坚实的基础。A four-parameter MDH kinematic model is established for the independently developed Ultra-T30 serial collaborative robot,the coordinates of calibration points are collected by FARO laser tracker,and a least squares objective function between the geometric parameter set of the robot and the end position error is constructed.Based on the joint offset identification of Levenberg-Marquardt algorithm,the geometric parameters of the robot are further compensated,and the calibration results are verified through 10 sets of validation points.The research results show that,after calibration,the average error of the robot Ultra-T30 is reduced by 79.67%,and the root-mean-square error is reduced by 79.75%compared to that before calibration,which lays a solid foundation for the commercial application and promotion of the UltraT30 robot.

关 键 词：机械设备智能终端联网 数字化矿山智能感知系统 AI大模型数据处理平台 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]