- Camera based grasping algorithms enable the handling of unknown objects without a complete CAD model. In some scenarios, the captured information from a single view is not sufficient or no grasp is possible. For these cases, the precise realignment of the gripper is difficult because a suitable rotation is part of an infinite solution space. In this paper, we propose a framework which automatically identifies correct rotations from point clouds to adjust the gripper. We validate our approach in a virtual environment for a parallel jaw gripper with multiple isolated and grouped industrial objects.
