基于标定和关节空间插值的工业机器人轨迹误差补偿

Compensation of Trajectory Error for Industrial Robots by Interpolation and Calibration Method in the Joint Space

轨迹精度是工业机器人重要的动态性能,目前工业机器人的轨迹精度远低于定位精度,提出一种基于机器人运动学标定和关节空间插值误差补偿的方法来提高机器人轨迹精度。基于MD-H方法建立机器人的运动学模型,在此基础上运用机器人微分运动学理论建立末端位置误差模型和轨迹误差模型。为克服最小二乘法等传统方法在数据噪声较大且不符合高斯分布时收敛慢甚至发散的问题,提出一种基于扩展卡尔曼滤波算法的机器人运动学参数辨识方法,实现运动学参数辨识的快速收敛。经过分析发现机器人误差在关节空间具有连续性的特点,为此提出一种关节空间插值误差补偿方法,建立网格形式的误差补偿数据库,并利用关节空间距离权重函数和已知的网格顶点误差计算各控制点的关节转角误差。通过试验对所提出的参数辨识和关节空间误差补偿方法进行了验证,试验结果表明:经过运动学参数辨识和补偿后机器人的绝对定位精度由1.039 mm提高到0.226mm,轨迹精度由2.532mm提高到1.873mm,应用关节空间插值误差补偿后机器人的轨迹精度进一步提高到1.464 mm。

Trajectory accuracy is an important dynamic performance of industrial robots,which is far lower than the positioning accuracy at present.A method based on kinematic parameter identification and interpolation error compensation in the joint space is proposed to improve the trajectory accuracy of industrial robots.The kinematics model of the robot is established based on MD-H method.Then,the positioning error model and the trajectory error model are established by using the differential kinematics theory of robots.To overcome the problem of slow convergence or even divergence of traditional methods such as the least square method when the data noise is large and does not conform to the Gaussian distribution,a robot kinematics parameter identification method based on extended Kalman filter(EKF) algorithm is proposed to realize the fast convergence of kinematic parameter identification.By analysis,we find that the error of robots in joint space has the characteristics of continuity.Therefore,an interpolation error compensation method in the joint space is proposed.The error compensation database in grid form is established.The joint angle error of each control point is calculated by using the distance weight function in the joint space.The experimental results show that the absolute positioning accuracy of the robot is increased from 1.039 mm to 0.226 mm,and the trajectory accuracy is increased from 2.532 mm to 1.873 mm.The trajectory accuracy of the robot is further improved to 1.464 mm after the interpolation error compensation in the joint space.