Regrasp

Regrasp is a term used in robotics and manipulation to describe the process of changing the grasp configuration of an object that is being held or manipulated by a robotic hand or gripper. The goal is to obtain a different grasp—often more stable, controllable, or better suited for the next task—without relying solely on the initial hold.

Two common approaches are in-hand regrasp and regrasp via an intermediate pose. In-hand regrasp relies on dexterous fingers and joint motion to rearrange the object's pose within the gripper, keeping contact with the object. Regrasp via an intermediate pose involves placing the object on a support surface, releasing and regrasping it from a different contact configuration, which can enable larger reorientations or grasps that are not possible in hand.

Planning regrasp sequences typically uses a regrasp graph or tree where nodes represent feasible grasps and edges represent motion or contact transitions. Planning methods include sampling-based approaches, optimization, and hybrid methods that combine in-hand manipulation with table-top placements. Real-world implementations must account for uncertainties in contact, grasp stability, collision avoidance, and sensing errors.

Applications span industrial manipulation, service robotics, and automated handling tasks, including assembly, packing, and object reorientation. Challenges include achieving reliable transitions, handling slender or slippery objects, and coordinating multi-fingered hands. Ongoing research aims to improve efficiency, robustness, and the integration of perception with grasp planning.