Uncalibrated Direct Visual Servoing Based on State Estimation

In order to cope with the low sampling rate of standard cameras and the time delay introduced by image processing, a predictor is constructed to estimate future positions of a moving target using BP neural network. The controller presented adopts adaptive PD control algorithm with uncertain gravity compensation and computes the demand torques for the robot based on current joint angles and joint angle rates. The direct visual servoing system is proved to be asymptotic stable with Lyapunov stability theory. The system does not require a robot kinematics model or a calibrated camera model. Simulation results show that this method provides a good steady-state tracking behaviour and keeps good robustness and adaptability at the same time. Additionally, the algorithm is very easy to be implemented with low computational complexity