Pick-and-Place

Picking an object up from one place, moving it, and setting it down at a target location — a basic manipulation task.

Pick-and-place is the most basic and most common manipulation task: grasp the target object, then move it to a target location and set it down. In industry it has long been used for sorting, loading and unloading, and palletizing, usually through fixed programs combined with visual localization. In robot learning, it is a standard test task in VLA and imitation-learning papers, with instructions typically of the form “put A into/onto B.” It looks simple, but it actually chains together several steps — recognizing the target, choosing a grasp point, planning a collision-free trajectory, and determining the placement pose — and it is often used as an atomic skill inside longer-horizon tasks. Google's Transporter Networks (2020) modeled pick-and-place as a spatial displacement, “pick from here, move to there,” and learned tasks such as stacking blocks and assembling kits, with sample efficiency several orders of magnitude better than the comparison methods of the time.

ExampleGiven the instruction “put the red block in the bowl,” the robot first grips the block, moves it above the bowl, then releases the gripper.

Also called

Pick and Place

Related

Grasping · Rearrangement · Skill Primitive · Sorting · Palletizing / Depalletizing · Transporter Networks

Sources

Transporter Networks: Rearranging the Visual World for Robotic Manipulation (arXiv:2010.14406)