Study On Pose Estimation Based On Hybrid Vision Of Robot Manipulator

With technological advances and the demands of social life,robots are widely used in industrial production and other fields of human activity,which is bound to make the robot undertake more complex operating tasks than industrial production and face a more complex working environment.Therefore,this requires the robot to have a stronger ability to obtain information from the outside world and to respond,so the visual servo comes into being.And pose estimation is an important research direction in the field of visual servo,is also a robot localization and navigation,target tracking and recognition,virtual reality,motion estimation and many other theoretical study and practice are the core problem to solve.First of all,this paper introduces the visual servo system of the mathematical model of camera and its geometric mapping relation,and the inverse kinematics analysis and modeling of the manipulator,The overall modeling and simulation of the Position-Based Visual Servo system is presented.Secondly,this paper studies pose estimation methods of based robotic visual servoing(RVS),and introduces the pose estimation algorithm based on kalman filter,that uses a new smooth variable structure filter(SVSF)pose estimation algorithm,It has good robustness and stability in the environment of model uncertainty and noise,and can significantly improve the robustness of estimation process.Considering the accuracy of the kalman filtering and the robustness of SVSF which is used based on kalman filtering and SVSF pose estimation algorithm named EK/UK-SVSF,combined with the advantages of both.Then,in view of the traditional pose estimation method based on monocular vision is vulnerable to camera calibration error,the influence of image noise and partial sheltering problem,this paper presents a multiple camera sensor fusion method,can effectively improve the accuracy and reliability of the robot pose computation.Method based on OWA(ordered weighted averaging)operator with EK/UK-SVSF pose estimation method,using decentralized data fusion structure to data fusion of image gathering information,and puts forward for multiple camera data fusion method of mechanical arm posture.This method is not limited by the number of robots and has strong expansibility.Simulation results show that the method presented in this paper has a high position posture calculation accuracy.