Uncalibrated visual servo algorithm based on RBF neural network optimizing maximum correntropy Kalman filter

Abstract

A radial basis function neural network (RBF) modified maximum correntropy Kalman filter algorithm (RBFMCKF algorithm) is proposed for robot uncalibrated visual servo positioning control. Because there may be non-Gaussian measurement noise in robot visual servo, the maximum correntropy criterion (MCC) is introduced into KF framework to suppress the impact of non-Gaussian noise on filtering accuracy, and then RBF neural network is used to adjusting the error produced by the maximum correntropy Kalman filter (MCKF) algorithm. The simulation results show that the proposed RBFMCKF is effective and can effectively suppress the influence of non Gaussian noise on the filtering accuracy. It provides an idea for further optimizing the robot’s uncalibrated visual servo positioning control.