Neural network based visual servo control for CNC load/unload manipulator

A visual servo control strategy based on fuzzy-neural networks is proposed for an eye-in-hand CNC load/unload manipulator in this paper. As visual servo control is an uncertain nonlinear strong coupling system, the real-time computation of feature jacobian matrix is very complicated, improving its poor real-time performance is a must. By approximating the mapping relationship between changes of target image features and robotic joints' positions with fuzzy-neural networks, which has the advantages of strong learning capability and fast learning speed, a novel controller is designed to achieve an effective operation for CNC load/unload manipulator. The following experiment result indicates that compared with BP and RBF neural network the proposed visual servo controller is of higher precision and convergence rate, enhancing the robust capability and accelerating the response time of the control system. (C) 2015 Published by Elsevier GmbH.