Pick and place is the manipulation task in which a robot grasps an object at one location and releases it at a defined target location, usually with a controlled position and orientation. It is the most common job for industrial and logistics robots, covering machine tending, packaging, kitting, palletizing, electronics assembly and order fulfillment.
In robot manipulation, pick and place is not one algorithm but a pipeline that chains perception, grasp planning, inverse kinematics, motion planning and gripper control. The quality of the whole operation is limited by its weakest link, which is why experienced integrators judge a cell by its cycle time, pick success rate and placement accuracy together.
How a pick and place cycle works
A single cycle follows a repeatable sequence. The robot locates the object, chooses a grasp, moves to a pre-grasp approach pose, descends, closes the gripper or activates vacuum, verifies the grasp, lifts and transfers the part, approaches the target, releases it and retreats. Each phase has its own speed limits, accuracy needs and failure checks.
| Phase | Main task | Typical check |
|---|---|---|
| Localization | Find object pose from fixtures or vision | Detection confidence, pose plausibility |
| Grasp selection | Choose gripper pose and contact points | Reachability, collision-free approach |
| Approach and grasp | Move to pre-grasp, then close slowly | Force, vacuum level or finger position |
| Transfer | Fast motion to the place area | Payload limits, part retention |
| Place and release | Position part and open gripper | Placement position, release confirmation |
Joint angles for each Cartesian pose come from inverse kinematics, while forward kinematics confirms where the tool actually is from the measured joint positions. Between poses, a motion planner or the robot controller generates smooth, time-optimal trajectories that respect joint velocity, acceleration and jerk limits.
Structured and unstructured picking
The difficulty of pick and place depends mostly on how much the robot knows in advance about where parts are. In a structured cell, fixtures, trays and feeders present every part at a known pose, so the robot can replay taught positions. In unstructured picking, parts lie randomly in bins or mixed totes, and vision must find each one.
| Scenario | Part presentation | Perception need | Main difficulty |
|---|---|---|---|
| Fixtured parts | Known pose in a jig or tray | None or simple presence check | Cycle time and repeatability |
| Conveyor tracking | Moving parts on a belt | 2D camera and encoder tracking | Synchronization with belt speed |
| Bin picking | Identical parts piled randomly | 3D vision and pose estimation | Occlusion, entanglement, collisions |
| Mixed-item picking | Many different products | Segmentation and grasp learning | Unknown shapes and materials |
Moving toward unstructured picking increases flexibility but lowers throughput and reliability, so many projects add simple mechanical presentation, such as vibratory feeders or flex feeders, to keep the vision problem easy.
Grippers and end effectors
The end effector often decides more of the result than the robot arm. Grasp planning chooses where and how to hold the object, but only within the possibilities that the gripper physically offers.
| End effector | Holding principle | Well suited to | Poorly suited to |
|---|---|---|---|
| Vacuum suction cup | Pressure difference on a sealed surface | Boxes, flat sheets, sealed packaging | Porous, perforated or very small parts |
| Parallel jaw gripper | Friction or form fit between two fingers | Machined parts, cylinders, rigid objects | Very soft or fragile items |
| Soft or adaptive gripper | Compliant fingers conforming to shape | Food, produce, irregular objects | High-precision placement |
| Magnetic gripper | Magnetic attraction | Steel sheets and ferrous parts | Nonferrous materials |
Grasp verification is essential at production speeds. Vacuum switches, finger position sensors and wrist force-torque readings detect missed or dropped parts within milliseconds, allowing the robot to retry instead of continuing an empty cycle.
Performance metrics and practical choices
Three numbers describe a pick and place system. Cycle time ranges from well under a second for delta robots on light parts to several seconds for vision-guided bin picking. Pick success rate is the share of attempts that end in a correct place, and placement accuracy is set by robot repeatability, gripper compliance and calibration between camera and robot.
Robot type follows the task. Delta robots dominate high-speed picking of light items from conveyors, SCARA robots serve fast planar assembly, six-axis arms handle reorientation and larger workspaces, and collaborative robots suit lower-speed tasks near people after a risk assessment.
Programming effort also differs widely. Fixtured cells are often taught by jogging the robot to a few reference poses, while vision-guided cells depend on a hand-eye calibration that relates camera coordinates to the robot base. Offline programming and simulation shorten commissioning by letting engineers check reach, cycle time and collisions before the cell is built, but the final positions are almost always touched up on the real hardware.
Common failures and how to prevent them
Field experience shows that most failures in pick and place have mechanical or calibration causes, not algorithmic ones.
- Hand-eye calibration error shifts every grasp by the same offset; recalibrate after any camera or tool change.
- Slipping parts during transfer come from excessive acceleration; reduce peak acceleration or improve grip force.
- Double picks of thin sheets need air knives, separation steps or thickness checks.
- Collisions with bin walls require approach vectors that account for the gripper volume, not only the tool point.
- Placement drift over time often reveals worn suction cups or loose fixtures rather than robot error.
A plain-language summary
Pick and place means exactly what it says: the robot picks something up and puts it somewhere else. It sounds simple, but the robot must first find the object, decide how to hold it, move without bumping into anything and set it down in the right spot.
Doing this fast and thousands of times without a mistake is the real challenge. That is why the gripper, the camera and careful setup matter as much as the robot arm itself.
An example from practice
A food packaging line produces chocolate bars that arrive on a conveyor in random positions. A delta robot with a vacuum gripper and a 2D camera above the belt picks each bar and places it into a moving tray, with the conveyor encoder letting the robot track bars as they travel.
Initially, bars occasionally rotated in the cups during fast transfers, which caused misplacements in the tray. Engineers softened acceleration at the end of each move and switched to a bellows cup with a better seal on the wrapper. The line then reached its target of well over one hundred picks per minute, with rejected bars caught by a vision check before sealing. Weekly inspection of the suction cups became part of maintenance, because worn cups were the main source of remaining drops.