Research On Uncalibrated Visual Servoing For Human-simulated Intelligent Manipulator Peg-in-hole Assembly

Manipulator peg-in-hole assembly is the common task in industrial automationassembly.Since vision sensors can obtain a wealth of information and non-contactmeasuring characteristics,manipulator peg-in-hole assembly based on vision has broadapplication foreground.Manipulator peg-in-hole assembly based on vision can be divided into twocategories including calibrated visual servoing and uncalibrated visual servoing.Theprocess of robot hand-eye calibration is tedious, and due to the model errors, noises ofmeasurementand system, the accurate calibration resultsare difficult to obtain. Inaddition, because calibration errors accumulate, or changes in the work environment,itis difficult to applycalibrated visual servoing to high precision industrial applicationssite in complex environment with strong interference.But, for image-based uncalibratedvisual servoing, the closed-loop controller is directly designed in image space by usingfeatureerrors. It is implemented by make camera and hand-eye relationships into theimage estimation ofJacobian matrix, where often refer to complex calculation,and needsto solve inverse of image Jacobian matrix in reality that might bring in singularityproblem. Consequently, visual servoing for manipulator peg-in-hole assembly is one ofhotspots and difficulty research issues in recent years.The direct estimation of image Jacobian matrix is avoided by imitate humanperception and identification in this paper. The method of uncalibrated visualservoingfor manipulator peg-in-holeassemblybased on Human-Simulated Intelligent isimplementedfrom the perspective of human simulation. The main research works asfollows:①The method of uncalibrated visual servoing for manipulatorpeg-in-holeassembly based on Human-Simulated Intelligent is proposed.The twoaxes parallel layout of binocular vision is employed, which every camerahas at least one axis parallel to the one axis of world coordinate system. It ensure somecoordinates approximately constant in camera coordinate system when manipulatortranslation along the axis in world coordinate system. The features of position decouplefrom orientationby chose point and line. The image features mapping model inqualitative and quantitative description is established. The method of uncalibrated visualservoing for manipulator peg-in-holeassemblybased on Human-Simulated Intelligentimage features mapping model is proposed.②The three-dimensional visual servoing simulation platform is designed, consistofthree-dimensionalscene graph module based on OpenInventor, manipulator controlmodule based on Orocos, human-machine interaction control module base onDirectInput, vision mapping module based on OpenCV and graphic user interfacemodule based on MFC. In this platform, the simulation of uncalibrated visual servoingfor manipulator peg-in-holeassemblybased on Human-Simulated Intelligent.③The binocular visual servoing experiment platform for6-axes manipulatorpeg-in-hole assembly is constructed, and the experiment of manipulator peg-in-holeassembly base on Human-Simulated Intelligent is made.The results of simulation and experiment demonstrate that the method proposed inthis paper can accomplish vision-guide peg-in-hole assembly task when the visionsystem parameters not calibrated and image Jacobian matrix not directly estimated.