Flexible Target Grasping Strategy of Industrial Robot Based on Eye-in-Hand 3D Machine Vision

Object grasping is the most common application direction of robots, and machine vision technology is also the most common application of grasping. With the deepening of the research of machine vision technology, the initial application in static target capture has gradually expanded to some moving target tracking and capture. Estimating and predicting the moving state of a target by detecting its position is a common method for obtaining a stable moving target. Once the target motion changes position due to unpredictable interference, the original predicted position will be invalid. Therefore, it is necessary to constantly locate and track the target and judge the deviation, so as to constantly approach the new target position and finally achieve the capture. In this paper, we utilize the open-source Robot Operating System (ROS) platform to investigate dynamic grasping strategies based on visual feedback. In this paper, we use ROS, based on UR5 robot arm and RealSense D435i camera, based on the eye-in-hand mode, to study the dynamic grasping strategy based on visual feedback. A visual feedback control strategy based on fixation approximation is designed to enable the grasping of moving targets on a conveyor belt. Finally, it is verified that the machine vision-driven feedback capturing algorithm substantially meets the practical application requirements of engineering, demonstrating high feasibility and accuracy.