Research On Some Intelligente Methods Applied In Robotic Visual Servoing System

Robot visual servoing is the fusion of results from many elemental areas including image procession, machine vision, control theory, robot kinematics and so on. The research on it has some challenge for its comprehensiveness. Some important factors infect the system performance include the high nonlinear, strong coupling, uncertain in the system and the demand of time-real in image process. Furthermore, intelligent methods and theory are developing quickly. They provide effective ways to solve the problems of nonlinear, time-real and inexactly models. This paper builds experiment system and simulating models for position-based and image-based visual servo systems. The problem of nonlinear, time-real and inexactly models are researched based on Genetic Algorithm (GA), Neural Network(NN) and Fuzzy Logic. The main jobs are as follows.According to the basic theory of target detection based on template matching and Genetic Algorithm(GA) image search, a new target detection method based on template zooming and multiple templates is proposed. Template zooming parameter and template type parameter are added to the GA searching operator. On the one hand, through dynamically building matching templates in the process of image searching, when the camera angle change is not big, the target with a change depth to the camera can be correctly detected. On the other hand, the matching templates can be automatically chosen through the optimization of the template type parameters. This makes easier to operate the visual servo system. The experiment results show that the affection and real-time of the proposed method can reach the demand of the visual process in the visual servo system.The kinematics analysis, camera calibrate, the inverse kinematics, and the visual servo of the robot RBT-6T/S04S are studied. Based on these jobs, a planar motive target tracking experiment system is realized, and is applied to robot catching fish experiment. Because the catching rate is not high as the result of the low intelligent level of the robot system, the fish escaping modes are analyzed and the move trajectories are predicted by NN. And a catching control strategy based on experience and NN prediction is proposed, which result is better than the former system. The proposed method improved the intelligent level of the catching system and the catching rate is clearly increased.The image Jocobian of RBT-6T/S04S robot is deduced. And a mathematic model of image-based visual servo system is built. Based on Robotics Toolbox (Release 7.1), a simulate method of space target catch with serial robot is proposed. And simulate experiments with proportion control are operated. An experiment system of image-based robot visual servo is constructed. Because that the template zooming proportion rate is changed with change of the visual depth, a method to estimate visual depth based on NN is proposed and is applied on target catching experiment with proportion control. The experiment result proves that the estimating method based on NN is effective.Six BP NN are built to reflect the relation of image feature change and robot joint change, the NN controller is realized and raised the response speed. But the experiment results also show that the effect of the NN controller is greatly connected with the robot study space. To enlarge the effective visual scope, according to the character of RBT-6T/S04S robot, a phased control strategy based on fuzzy logic and NN is proposed. The robot control process is divided into remote control phase used fuzzy logic and near control phase used NN controller. This solves the problem of the small control area in NN controller and the difficulty in determining the fuzzy formula. The relative experiment results show that this phased control strategy can accommodate the high nonlinear, strong coupling, uncertain in the robot visual servo system. And this system has a bigger control area, better accuracy and a more quickly response speed. The proposed phased control strategy provides a new approach to the application of the intelligent method on robot visual servo system.