Adaptive Robust Kalman Filter for Vision-based Pose Estimation of Industrial Robots

Abstract

In this paper, an adaptive robust Kalman filter (ARKF) for precise and robust pose detection of industrial robots is presented. The proposed ARKF exploits the advantages of adaptive estimation method for states noise covariance (Q), least square identification for measurement noise covariance (R) and a robust mechanism for state variables error covariance (P). In simulation on PUMA 560, the comparison between the proposed ARKF and other well-known version of Kalman filter such as adaptive Kalman filter (AKF) and standard Kalman filter (SKF) shows the superiority of the ARKF in terms of root mean square (RMS) and Variance (Var) of filtered errors. The ARKF outperforms above-mentioned methods both in smooth filtering and in signal tracking. Simulation results reveal the superior tracking performance of the ARKF when the robot is subjected to the measurement noises and uncertainties.