基于IRLS算法的机器人动力学参数辨识

Identification of robot dynamic parameters based on IRLS algorithm

为提高机器人动力学参数辨识的准确性,提出了一种基于迭代加权最小二乘(Iterative Reweighted Least Squares,IRLS)算法的辨识方法。首先推导了机器人的线性动力学模型,随后提出了一种改进摩擦模型,并设计了改进傅里叶级数作为激励轨迹采集数据。为提升动力学参数辨识的准确性,在加权最小二乘法基础上进行改进,提出了IRLS算法对动力学参数进行辨识。最后以六自由度机器人为试验对象,进行了参数辨识试验。结果表明,基于IRLS算法的辨识方法与加权最小二乘法相比,前3个关节力矩误差的均方根(Root Mean Square,RMS)值降低了13.28%,后3个关节力矩误差的RMS值降低了28.57%,6个关节力矩误差的RMS值平均降低了17.15%,证明了基于IRLS算法的辨识方法的有效性。

In order to improve the accuracy of robot dynamic parameter identification,an identification method based on IRLS algorithm was proposed.Firstly,the linear dynamic model of the robot was derived,and then an improved friction model was proposed,and the improved Fourier series was designed as the excitation trajectory to collect data.In order to improve the accuracy of dynamic parameter identification,IRLS algorithm was proposed based on weighted least square method.Finally,taking the 6-DOF robot as the experimental object,the parameter identification test was carried out.The results show that compared with the weighted least square method,the RMS of the torque error of the first three joints is reduced by 13.28 %,the RMS of the torque error of the last three joints is reduced by 28.57 %,and the RMS of the torque error of six joints is reduced by 17.15 %,which proves the effectiveness of the identification method based on IRLS algorithm.