Reorientating objects with a gripping hand and a table surface

We present a robotic system that can perform difficult reorientation tasks like flipping. The system uses a gripping hand and takes the advantages of external surfaces. It is composed by three components. In grasp planning, the system finds the grasps that fulfill force closure and resist high external wrenches. In placement planning, the system computes the stable placements of an object on the table surface and their associated grasps. In manipulation planning, the system builds a regrasp graph by connecting the stable placements and lazily searches the regrasp graph to find a reorientation sequence. The work is evaluated by thousands of simulation experiments and demonstrated by real-world executions. With the help of the presented implementation details, we physically validate the utility of an external table surface.