Research About Manipulator Visual Servoing System Based On RBF Neural Network

The procedure of the manipulator visual servo system is to handle the image information which is got from the vision sensor, and we take the handled information as feedback signal of the whole system, the manipulator’s closed loop control system is then built. Since entering the new century, the industrial automation degree is higher and higher along with technology developing. The industrial automation degree mostly relies on the application of robot in industry. And the application of vision information helps a manipulator sense the environment around it. So the study of manipulator visual servoing system attracts a lot of researchers’ attention. In this article, we do some research on manipulator visual servoing system, the main work is as following:First of all, the RBF neural network is adopted to approach the inertia matrix, centrifugal force term and the gravity term, the three terms are usually used to describe the manipulator system. The method solves the problem of manipulator system’s parameter variety, and the parameter’s variance is resulted from the linkage’s uncertainty or the disturbance of the environment that surround the manipulator. A dynamic controller is put forward based on the manipulator dynamic equation, and the Lyapunov stability theory is used to certificate the controller’s stability. And the controller is used on the manipulator visual system. Build the simulink model in Matlab to certificate its efficiency, and the results of the simulink model certificate that the controller is effective.Secondly, due to the coupling characteristics and the nonlinearity in manipulator, it exits some uncertainty, which can’t be ignored while controlling the manipulator. So a RBF neural network is proposed to identify the uncertainty, and compensate the result in the proposed controller. Take a two-linkage manipulator as the research model, propose the kinematic controller and the dynamic controller. Take the error of the desired feature point and the real-time feature point as the input of the kinematic controller, and the output of this controller is the joint angle velocity of the manipulator. Take the joint angle velocity as the dynamic controller’s input, and compensate the identified uncertainty in this dynamic controller, the output of dynamic controller is the torque signal that drives the manipulator. Under the effect of this two controllers, the manipulator can reach the desired point, that is to say, it has completed the servoing task successfully.Finally, the system’s simulation module is built in Simulink environment of MATLAB to certificate the effectiveness of the proposed controllers. The final result certificates that the controllers are effective.