Recursive Calibration of Industrial Manipulators by Adaptive Filtering

Abstract

The calibration scheme of robot forward kinematics presented in this paper has a number of features. Firstly, robot kinematic errors are modeled in a recursive format and as such, the number of measurements that need to be taken for calibration can be determined by studying the rate of convergence of estimation error covariance. Secondly, a simplified adaptive filtering algorithm is used to deal with unknown measurement noise statistics and unknown robot motion repeatability characteristics in estimating the kinematic errors. Thirdly, a laser interferometry system is used to measure positions of a robot end-effector in world coordinates. The measurement system was implemented in experiments involving a three degree-of-freedom gantry robot. The adaptive filtering of the experimental data identified 0.5 to 1.5 percent errors in representative kinematic parameters of the given robot by taking into account measurement noise and robot repeatability.