Kernel machines for uncalibrated visual servoing of robots

A new modelling method of image Jacobian estimation is presented for uncalibrated visual servoing of robots, in which a kernel recursive least squares (KRLS) technique is used for non-linear mapping between target image features and robot joint angles, and an image Jacobian expression is derived from the KRLS algorithm with gaussian kernel. The simulations of robot visual servoing with eye-in-hand camera configuration are conducted using the KRLS Jacobian estimator and the same are compared with SVR and LS-SVM Jacobian estimators. The simulation results have shown that the robot visual servoing converges at the desired goal and KRLS is proved to be a better choice.