Pick-and-Place
- Also called: pick and place, pick-place
- Category: Planning & manipulation › Manipulation
- Tags: manipulation, grasping, industrial-automation, benchmark-task, robot-learning
The manipulation task of grasping an object at one location and releasing it at a specified target pose; the basic operation of industrial automation and a standard robot learning benchmark.
Definition
Pick-and-place is the task of moving an object from where it is to where it should be: the robot approaches the object, grasps it, lifts and transports it, and releases it at a target position and orientation. It is the most common job given to industrial robots, from SCARA and delta robots moving parts on conveyor lines to dedicated machines that place electronic components on circuit boards, and it is also the most common task family in robot learning datasets and benchmarks.
A complete pick-and-place system chains several problems. Perception locates the object and, when needed, estimates its pose. Grasp planning chooses where and how to hold it, whether with an antipodal parallel-jaw grasp, a suction cup or a multi-fingered hand. Motion planning produces collision-free approach, lift, transport and retreat motions, usually with straight-line segments near the object. Placement must find a pose where the object is stable, does not collide with its surroundings and satisfies the task, for example upright, inserted into a fixture or stacked. In structured factories, fixtures and known part geometry remove most of the uncertainty, so the task reduces to taught positions executed at high speed. In unstructured settings such as warehouse bin picking or household tidying, the hard parts are clutter, unknown objects, transparent or deformable items and occlusion.
In learning research, pick-and-place episodes are often specified by language, such as placing an object in a named container, and are used to test generalization to new objects, scenes and instructions. A grasp that succeeds is not the same as a task that succeeds: the object must be held in a way that allows the placement, which is why grasp choice is often conditioned on the place pose.
Why it matters
Most practical manipulation reduces to picking things up and putting them somewhere. Understanding the full chain of perception, grasping, motion planning and placement, and where each one fails, is basic working knowledge for anyone building or evaluating a manipulation system.
Key applications
- Machine tending and parts transfer on production lines
- Electronics assembly with surface-mount component placement
- Warehouse order picking and bin picking
- Packaging and palletizing
- Language-conditioned manipulation benchmarks for vision-language-action models
- Household tidying and assistive fetching
In practice
A logistics cell uses a six-axis arm with a suction gripper above a tote of mixed items. A depth camera images the tote, a model proposes suction points on flat, exposed surfaces, and each candidate is scored for seal quality and collision risk with the tote walls. The arm approaches along the surface normal, checks vacuum pressure to confirm the seal, lifts vertically to clear neighboring items and moves to a shipping box. The place pose is computed from the box's free space so the item drops only a short distance. Failed picks, detected by loss of vacuum, are logged with the depth image so the grasp model can be retrained on its mistakes.
Common misconceptions
- Pick-and-place is a solved problem — it is routine with fixtured, known parts, but clutter, unknown, transparent and deformable objects keep it an active research area
- Placing is the easy half — placement needs a stable, collision-free pose that meets the task, and some grasps make the required placement impossible
- A successful grasp means a successful task — objects can slip, rotate in the gripper or be held in a way that blocks the intended placement
- High success rates on a benchmark mean real-world readiness — benchmark scenes and object sets are narrow, so results often fail to transfer to new objects or lighting
Who should know this
- Industrial automation and robot cell integrators
- Manipulation and grasp planning engineers
- Robot learning researchers evaluating policies
- Makers building their first arm projects
Related technologies
- Vacuum and Suction Grippers
- Parallel-Jaw Grippers
- MoveIt
- Delta and SCARA Robots
- Object Pose Estimation
Related terms
Related projects
- LeRobot: HuggingFace ML library for real-world robotics in PyTorch
- OpenPI: Physical Intelligence VLA foundation models for general robot control
- Isaac Lab: NVIDIA robot learning framework: one task API across Newton, PhysX and Isaac Sim
- NVIDIA Isaac GR00T: NVIDIA open foundation model for humanoid robot reasoning and skills
- OpenVLA: 7B-parameter Vision-Language-Action model for robotic manipulation
- Drake: MIT robotics toolbox for model-based dynamics, optimization, and control system design