Robot manipulation using image-based visual servoing control with robust state estimation

摘要

In unstructured environment, this paper presents an image-based visual servoing control approach with robust state estimation for robot manipulation. The image Jacobian on-line identification problems are firstly addressed by introducing neural network (NN) aid Kalman filtering (KF) schema. The neural network plays exactly the role of the error estimator, to compensate the state-estimation-errors of KF. Then the proposed image-based visual servoing control approach has guaranteed the robustness with respect to destabilized system attached dynamic noises. Furthermore, the presented approach without requiring the intrinsic and extrinsic parameters of the camera, also the hand-eye do without calibrated during robot manipulation. To demonstrate the validity and practicality of proposed image-based visual servoing approach, various robot positioning experiment results have been presented using a six-degree-of-freedom robotic manipulator with eye-in-hand configurations.