Vision-Guided Grasping of a Strut for Truss Structure Assembly

Abstract

Monocular information from a gripper-mounted camera is used to sew0 a robot gripper to grasp a strut. The procedure is divided into four phases: learn, recognition, alignment, and approach. In the learn phase, a strut is placed in the gripper and the pose estimate from the camera is used as a servo target. The recognition phase verifies the presence of a cylinder in the camera field of view. The alignment phase processes only selected scan regions in the image. Relative motion information a5 then used to generate an extrapolated posebased trajectory to guide the gripper to a certain dastance from the strut. The approach phase uses the last pose estimate and “blindly” moves to grasp the strut. By using a sample feducial stripe on the strut, rapid and robust vision-guided grasping can be achieved.