Research On Object Tracking And Dead-zone Compensation For Robot Visual Servoing Control

Vision is an important means for robot to get the information of environment and isalso an important guarantee for improving the adapability of robot．With the rapiddevelopment of the visual measurement and control of the robot，the research of robotvisual servo control has made new progress，and the research results have been moreand more widely used in the fields of industrial robots，military，aeronautics and thespace exploration．Therefore，the visual servo control plays an important role in thefield of robot，and the algorithm of moving object detection and control of robot hasbeen paid more attention by researchers in robotics．The subject of this thesis is visual control for robots．Firstly，according to theknowledge about theory of robot and vision measurement，the kinematics equationsof manipulator is builded and the homogeneous transformation matrix between the endcoordinate of robots and the camera coordinate is obtain．Then the homogeneoustransformation matrix between base frame of robots and the camera coordinate isobtain．The inverse kinematics is introduced in this thesis．Secondly，by the needsof getting moving target’s location obtained by vision measurement，two improvedalgorithms for detecting moving targets are designed， based on the existingalgorithms of detections about moving targets．In order to control the manipulator totrack the trajectory obtained by the vision measurement，the model of the manipulatordynamics is established on the basis of fully considering dead-zone character andunknown disturbance，when analysis the robot dynamics．By the theories of slidingmode control and adaptive control，three control schemes are designed．The validityof the control schemes is proved by using theoretical analysis，Matlab simulation andtests．Lastly，on the basis of research，the scheme which is based on position basedvisual servo control for tracking moving target is designed，the feasibility of thescheme is discussed in theory．The experimental results show that two improved algorithms for detecting moving targets are able to generate background image and achieve target detection in the scenewhich exists the variable factors of moving objects，the variation of light etc， andlocation of target is obtained by vision measurement．The control schemes designedfor manipulator of dead-zone and uncertainty disturbance can achieve good control ofmanipulator．...