Study On Object Grasping Of Mobile Manipulator Based On Visual Servoing

In the smart factory of Industry 4.0,mobile manipulator has been widely used because of its flexibility,large work space and high level of intelligence.In the conveyor workstation,mobile manipulator needs to track and grasp the workpiece moving on the conveyor.In this situation,different angle deviations between the object pose and the moving base pose need to be considered.In the human-machine cooperation situation,mobile manipulator needs to grasp objects which are passed from human's hands.After the grasping operation,manipulator needs to place the grasped object on the platform of the moving base.The traditional solution for conveyor picking problem in industry includes high speed cameras and it only consider the grasping in twodimensional plane.Besides,when an ordinary camera is used,there remains two problems: time delay of the vision system and the system response speed problem due to the low frequency of getting images.About the hardware system,the mobile manipulator is equipped with two cameras.The hand-eye camera is used to track and pick the target object in the grasping system based on visual servoing.The onvehicle camera is used to model the obstacles in the placing system based on motion planning.About the control system,the software structure of the grasping system are designed based on visual servoing and joint space velocity control of the manipulator.About the grasping problem,PID plus target velocity feedforward controller and object tracking algorithms based on moving-edges and Kanade-Lucas-Tomasi keypoints tracking are proposed.Firstly the jacobian matrix of a 6-DOF manipulator is deduced.Then the approach that using PID and joint space velocity control to realize the position and pose tracking of the target object is proposed.After that,interpolation and prediction based on Kalman filter are used to solve the time delay problem and the low frequency problem of getting images for the vision system.About the placing problem after grasping,the rapidly-exploring random tree motion planning algorithm with the environmental obstacles modelling by the depth camera is adopted to acquire the collision-free path during the placing process.A simulation platform is established.So the control framework of the placing system based on motion planning is proposed.In the conveyor grasping situation,the moving base pose and the object pose on the conveyor are uncertain,mathematical modeling is done and the grasping scheme is determined for general case.In the human hand grasping situation,grasping is divided into two processes: tracking the pre-grasping point and approaching.Finally,experiments are conducted to verify the effectiveness of the algorithm.