Research On Robotic Arm Visual Servoing Control Based On Image Feature

The IBVS(image based visual servoing) without uncalibrated control method avoid the problem of position estimation, inverse kinematic and camera parameter calibration, and has the properties of less calculation, high accuracy, design flexibility and other advantages. It can realize the arm control with an estimation value of camera parameter.Firstly, the basic theory of visual servo is introduced, and the uncalibrated visual servo method and its latest development are discussed in detail. Subsequently, the manipulator forword-kinematic and invese-kinematic are discussed, and camera calibration method and the problems existed are introduced. Then,the image Jacobian matrix for six degrees of freedom robotic arm is derived based on point features. Online estimation image Jacobian matrix is an important part of uncalibrated visual servo method, a control method based on quasi-Newton method and Jacobian online estimation methods based on dynamic Broyden method for six degrees of freedom, eye-in-hand visual servoing system are derived. Then, in order to improve the global performance, this paper presents a control method based on trustregion which can achieve a wide range of asymptotically stable without precise calibration values. In order to reduce the influence of image noise, an online-estimation algorithm based on Kalman filter are discussed, and an adaptive Kalman filter also be used to estimate Jacobian without any prior knowledge of system noise. And simulation based on MATLAB verify the validity of all above algorithm.Image-based visual servo control ensure the performace in the image plane, but does not guarantee the optimal performance in Cartesian space, this will leads to the camera retreat phenomenon, even servo failure. The partitioned method which decouples the Z-axis movements from other axis by image Jacobian and the HVS(hybrid visual servoing) method decouples the rotation and translation movements by homography matrix are discussed. Both two algorithm have good performance in image space and Caresian Space. The effectiveness of the above algorithm is validated by MATLAB simulation.Finally, in order to apply the algorithm for the practical scene, two method for Homography Matrix are validated, and the IBVS and HVS method are verified on a Viper850 Robotic arm based on a C++ visual servoing platform.