Uncalibrated visual servoing based on Kalman filter and mixed-kernel online sequential extreme learning machine for robot manipulator

Visual servoing systems may suffer from interference by system noise when a Kalman filter is used to obtain a Jacobian matrix. Such interference may result in slow and poor convergence performance of the servoing system. To overcome these problems, we propose a mixed-kernel online sequential extreme learning machine (MIXEDKOSELM) with Kalman filter, which corrects the error of Kalman filtering algorithm, thus improving the accuracy of the image-based visual servoing (IBVS) system significantly. The proposed KF-MIXEDKOSELM-IBVS does not require the camera parameters in the servoing process, and it is highly robust to disturbance and noise statistical errors. The proposed KF-MIXEDKOSELM-IBVS is validated using the PUMA 560 manipulator in the MATLAB simulation environment. The simulation results clearly reveal that the KF-MIXEDKOSELM-IBVS algorithm has excellent performance by being robust and accurate.