Visual Servoing Of A6-DOF Manipulator Based On Human-simulated Intelligent Control

By combining visual perception technology with motion control of manipulator,visual servoing is applied to collect unstructured environment information and achievevisual positioning and tracking of manipulator under the condition of non-contact,which expands the applicationrange of manipulator, such as the application ofpeg-in-hole assembly and other industrial production.Without accurate parameters of manipulator kinematics model and camera model,the controller of uncalibrated visual servoing system is designed by the errors of imagefeaturesto realize the visual positioning and tracking control, additionally, uncalibratedvisual servoing owns stronger robustness for the serious non-linear existence anduncertainty of complicated disturbance in Cartesian space. However, considering theabsence of visual mapping model and the strong coupling between the manipulatormotion in Cartesian space and image features motion projected in the image planar, thevisual servoing is essentially a model-unknown strong coupling MIMO system, whichis the research focus and emphasis of robotic technology.Humanoid-Simulated Intelligent Control (HSIC) theory designs the controller byimitating the control behavior and architecture of human, and the main goal is not theprecise model of the controlled objects, which is especially effective for complexnon-linear problem solving. This paper based on HSIC studies the achievement of thetask of peg-in-hole assembly and uncalibrated visual servoing control of six DOF(degree of freedom) manipulator by imitating the character of unknown or inaccuratemodel of human behavior. Additionally, the intelligent control method of nonlinearproblem of uncalibrated visual servoing with unknown dynamic model is alsoresearched. The main contents of the paper are organized as follows:①The control method based on HSIC is proposed to acheive visual servoing of6dof manipulator. In this method, the point feature is selected to establish image featuremodel qualitatively. By fixing image Jacobian and online image Jacobian, the hand-eyerelationship is approximate. Thus the errors between desired features and feedbackfeatures in image planar are projected to manipulator walking space. Then we establishthe qualitative and quantitative characteristics model of the system by the priorknowledge of manipulator and camera. Finally, the control problem is solved by thetwo-order mapping relation in qualitative decision and quantitative control. ②Based on the models of camera and manipulator, a simulation platform is builtfor3DOF and5DOF visual positioning experiments in the manipulator working space.③An experimental platform with fixed binocular vision is built based on6DOFmanipulator DENSO, and visual servoing experiments are carried out with unknownboth parameters of manipulator kinematics and camera model.The validity of controller designed by HSIC is proved by the3DOF and5DOFvisual positioning simulation and experiment of manipulator, and the method based onHSIC is effective to control an unmodeled dynamic nonlinear system.