Research On Uncalibrated Visual Servo System Of Industrial Manipulator

In recent years,with the rapid development of machine vision and robot technology,the use of visual information for robot motion control to complete the servo task has become a hot spot in the field of robot research.Visual servo is to use the characteristic information of the collected image as the feedback signal of the system,calculate the change of joint angle and adjust the end effector to the target pose,then carry out image acquisition,cycle calculation and adjustment operation,and finally complete the motion task.The traditional visual servo system needs system and model calibration,and the error of calibration process will affect the control accuracy of the system,so the research of uncalibrated visual servo system is introduced.The uncalibrated visual servo system does not need to obtain the calibration information of the system and camera model in advance.The objective function of the system is constructed by using the difference between the image features of the current image and the desired image,and an appropriate controller is designed.By online estimating the image Jacobian matrix,the change of the joint angle is obtained,and the end effector is adjusted to make the object function converge to a allowable error range,and then complete the movement task.The main research work of this paper includes:(1)This paper investigates the background and significance of industrial robot visual servo research,summarizes the status quo of uncalibrated visual servo system at home and abroad,classifies and compares the current visual servo system,and points out that this paper studies uncalibrated visual servo system based on monocular "Eye-in-hand".(2)The paper explains the principle knowledge of manipulator kinematics and visual servo,and proposes a system model of uncalibrated visual servo research based on single eye in hand.The research of the model includes extracting image features,constructing objective function,designing motion controller,image Jacobian matrix online estimation and so on,the innovation of this paper is to construct the object function of eight parameters and introduce an optimal particle filter algorithm for on-line estimation of image Jacobian matrix.(3)According to the feature extraction,the image feature region is obtained by extracting color feature,and the image feature information is calculated and the object function is constructed by using the information;Aiming at constructing the object function,the eight parameter method of image feature formation is added based on the existing six parameter method,which increases the average area and the average perimeter of the image feature area,and uses this method to construct the object function of uncalibrated visual servo system;For the design of motion control law,the nonlinear variance minimization is adopted to design the controller,and the demand of Jacobian matrix online estimation is derived;For the image Jacobian online estimation algorithm,the existence of non Gaussian noise affects the accuracy of Kalman filter online estimation.Particle filter algorithm has the ability to deal with non Gaussian noise,but the computation is large.Therefore,this paper introduces a shrimp swarm optimization particle filter algorithm to on-line estimate Jacobian matrix.(4)Combined with the robot control system,using QT and other software to design the interface,the uncalibrated visual servo system based on "Eye-in-hand" is built to realize the visual servo task of two-dimensional plane object positioning.