基于GBDT算法的机器人定位误差分级补偿方法  

Graded Compensation Method of Robot Positioning Error Based on GBDT Algorithm

作　　者：李晓昆 叶伯生[1] 邵柏岩 金雄程 李思澳 黎晗 LI Xiaokun;YE Bosheng;SHAO Baiyan;JIN Xiongcheng;LI Siao;LI Han(School of Mechanical Science and Engineering,Huazhong University of Science and Technology,Wuhan Hubei 430074,China)

机构地区：[1]华中科技大学机械科学与工程学院,湖北武汉430074

出　　处：《机床与液压》2024年第11期1-6,共6页Machine Tool & Hydraulics

基　　金：湖北省重点研发计划项目(2021BAA197)。

摘　　要：为进一步提高工业机器人的定位精度,提出一种分级补偿的方法以降低几何和非几何因素引起的定位误差。使用遗传算法优化最小二乘法(GA-LS)进行几何参数误差辨识并补偿到机器人运动学模型中,再通过梯度提升树(GBDT)算法对残余非几何参数误差进行预测,并对残余误差进行补偿,最后以UR10机器人为研究对象进行了实验,验证该方法的准确性。实验结果表明:此分级补偿方法能有效提高机器人的绝对定位精度,补偿后机器人的平均定位误差由2.381 mm降低至0.156 mm,定位精度提升了93.4%;均方根定位误差由2.417 mm降低至0.163 mm,定位精度提升了93.2%。实验结果验证了此分级补偿方法的有效性。In order to further improve the positioning accuracy of industrial robots,a graded compensation method was proposed to reduce the positioning error caused by geometric and non-geometric factors.The genetic algorithm optimized least squares method(GA-LS) was used to identify the geometric parameter errors,and then the geometric parameter errors were compensated it into the robot kinematics model.Then the gradient boosting decision tree(GBDT) model was used to predict and compensate the residual non-geometric parameter errors,and finally the UR10 robot was used as the research object for experiments to verify the accuracy of the method.The experimental results show that the graded compensation method can effectively improve the absolute positioning accuracy of the robot,and the average positioning error of the robot is reduced from 2.381 mm to 0.156 mm after compensation,the positioning accuracy is increased by 93.4%,the root mean square positioning error is reduced from 2.417 mm to 0.163 mm,and the positioning accuracy is improved by 93.2%.The effectiveness of the graded compensation method is verified by the experimental results.

关 键 词：机器人标定 误差辨识 绝对定位精度 梯度提升树 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]