Research On Hand Eye Coordination Method For Industrial Grasp Application

With the continuous integration of industrialization and information technology,the intelligent industry,represented by robots,is booming.The application of the robot also extends from the traditional automobile manufacturing and mechanical processing to the 3C,logistics and other industries.At present,some domestic electronic manufacturers are competing for industrial upgrading,and put forward"machine replacement".How to achieve flexible and accurate grasping of robots instead of manual work has become an urgent problem to be solved.In this paper,the problem of accurate grasping of objects by the hand eye coordination is explored and studied.The main contents and results of the research include the following aspects:1.we designed a system based ROS platform for hand eye coordination to achieve the object precise grasp.We proposed an improved algorithm based on open source planning algorithm and the algorithm of object location point calibration.The planning algorithm adds constraints to the original planning results,improves the smoothness of the motion process,and inserts the middle point appropriately in complex environment,so as to improve the planning efficiency.By locating the object's grasping point,we can determine the pose of robot's coordinate system,and then get the robot's target pose,so that it can achieve precise grasp.2.mobile grasping,that is,when the robot's base is moved to the location near the operation table,robot requires rapid deployment and implementation of grasping operation.In this paper,a servo method based on iterative adjustment is proposed.in view of the movement of the robot base,a servo method based on iterative adjustment is proposed in this paper.The method divides the grasp into two steps,that is,the pre pgrasping state and the grasping state.The coordinate transformation from pre grasping state to grasping state is set as a constant.The core problem becomes how to reach the pre grasping state.Such a method can be solved without a calibration tool coordinate system,but by a teaching method.The pre-grasping state is reached by the iterative adjustment method,which greatly improves the positioning accuracy.3.different pose estimation methods should be adopted for different objects,and a series of calibration tasks should be carried out.In this case,a servo method based on deep learning is proposed in this paper.Similar to the method in 2,the grasp is still divided into two steps.A mapping relationship from the pose of the image in place is established.The pose in the pre grasping state is taken as the initial pose,and the offset relative to the initial pose is taken as an absolute pose.The PoseNet network structure is trained.This method simplifies the operation process,achieves high precision of grasping,has a certain robustness to light intensity and can move away from the camera's field of view.