Vision Servo Control Of Manipulator Based On Model-free Adaptive Predictive Control

With the continuous progress of computer vision,visual servoing technology is introduced into robot control and becomes one of the core research contents in the field of robot control.The robot obtains image information through the camera in the process of working,which makes the robot operating system more flexible.Therefore,the research on the visual servo control of the manipulator is meaningful.For the multiple-input and multiple-output(MIMO)nonlinear system like the visual servo system of a 6-DOF manipulator,a visual servo controller based on Model Free Adaptive Control(MFAC)algorithm with uncertain parameters are designed in this thesis.The main researches are as follows.(1)Considering that the results of image processing directly affect the control precision,the methods of image preprocessing,edge detection and feature point extraction are studied.A method to determine pixel coordinates of feature points in target image based on image processing is presented.(2)A vision servo control method based on model-free adaptive control is proposed and applied to uncertain model system.The difference between the feature point information of the target object and the desired feature is used as the feedback information to design the controller to drive the manipulator.The simulation results show that the model-free adaptive vision control method is effective in the case of model uncertainty of the visual servo system.(3)In order to improve the control precision and rate of convergence,this paper proposed a model-free adaptive predictive control method(MFAPC)for visual servo system,which combine model free adaptive control and predictive control.By establishing equivalent linearized dynamic model at each moment,using the input and output data of the controlled system to estimate the pseudo partial derivatives of the system,the optimal control quantity is calculated by the predictive control algorithm.The simulation results show that the model-free adaptive visual predictive control method is effective in the environment with no calibration and the unknown model condition.