Research On Control Of Image-based Robot Binocular Visual Servo System

People pay close attention to the robot research with the visual function along withthe rapid development of the robot and computer technology. This paper mainlyresearches on control of image-based robot visual servo system, which applies a binocularvisual model with the depth information. The robot systems can accomplish a positioningfunction or a positioning origin function for a static object. The main research works aresummarized as follows:Firstly, the parameter vector of binocular visual model contains depth information, sothe object depth needs not to be known or estimated when designing controllers. Theimage-based direct servo PD controller is presented for eye-in-hand robot dynamics,which applies robot Jacobian matrix transpose. The global asymptotical stability of thesystem is proved by Lyapunov and Salle theorems.Secondly, for the binocular vision servo system model, a global stabilized controlleris been designed based on the properties of weak control Lyapunov functions, without thereal time estimation of image Jacobian matrix and its inverse operation. The controller canmake an arbitrary target point of the field back to the origin.Thirdly, the combined controller with PI controller and BP neural network controlleris been presented for image-based robot visual servo system. PI controller outputs adesired robot joint velocity according to the image information error, then, BP neuralnetwork controller computes the joint torque to drive the robot to a desired position. Thesystem stability is proved by Lyapunov theorems.Finally, the controllers of this paper are simulated by Matlab/Simulink. Thecontrollers are effective through the objective analysis of the simulation results.