Knowledge · 3D bin picking
What is 3D bin picking and how does it work?
3D bin picking automates the repetitive task of picking parts that lie in random positions inside a container. A vision system maps the visible scene in three dimensions, after which software and robot work together to execute a reachable, safe grasp.
From manual work to a controlled process
Why manufacturers automate bin picking
Picking loose parts from a container is repetitive work. An operator repeatedly has to inspect the contents, find a usable part, choose a grasp and present the part correctly to the next process. Ergonomics also matter when parts are heavy, sharp or difficult to reach.
Automation can reduce reliance on this manual work and feed a machine or assembly step more consistently. Whether it succeeds in production does not depend on the camera alone. Part geometry, fill level, gripper, robot reach, required capacity and recovery after a failed pick form one system.
The process
From 3D capture to a confirmed pick
A robust cell evaluates more than the uppermost outline. Every step must be sufficiently reliable and fast for the production process.
- 01
Capture the contents
The 3D camera measures visible surfaces and returns depth data. Parts underneath other parts are not visible at that moment.
- 02
Find a usable part
Software decides which visible areas can be used. Depending on the application, it estimates an object pose or searches directly for suitable grasp surfaces.
- 03
Select a grasp
The system evaluates possible contact points against the gripper, the required presentation and known restrictions of the part.
- 04
Check reach and collisions
The selected robot pose must be reachable without striking the container, neighbouring parts or cell hardware.
- 05
Pick and place
The robot approaches, grips and moves the part to the agreed position for machining, assembly or further handling.
- 06
Verify the result
A vacuum switch, gripper signal or additional vision check confirms the pick. A predefined recovery action follows when something fails.
Use no more technology than necessary
Not every feeding task requires full bin picking
When parts lie flat or can be aligned mechanically, 2D vision may be enough. 3D robot guidance can suit variable height or orientation. Full bin picking becomes relevant when parts are randomly interlocked and the system must calculate a new grasp and motion for each cycle.
Read the decision guide: when is 3D bin picking needed? →What production reliability depends on
- Visibility: parts may partially hide one another; a new capture must reveal viable candidates.
- Gripper: access, seal, force and part variation determine which grasps are executable.
- Calibration: camera, robot and tool coordinates must demonstrably agree.
- Recovery: the cell must know what to do when no candidate is found, a part is lost or placement fails.
Business feasibility
Calculate with the complete process, not one ideal pick
Start with required production capacity: available production time divided by the number of correct picks required. Compare that target cycle time with measurements from representative containers, including rescans, rejected grasps, recovery actions and the final layer.
The business case should also account for reduced manual lifting and reaching, operator availability, downtime of the machine being fed, changeover and maintenance. This shows which problem the investment must solve and which assumptions still need testing.
Explore each question
From overview to technical choice
Cycle time
Combine sensing, computation, gripping, motion and recovery in one time budget.
Estimate position and orientation
See how a 6D pose is calculated and how measurement errors reach the robot.
Choose an executable grasp
Move from possible contact points to a reachable, collision-free grasp.
2D or 3D robot vision
Compare what each approach measures and which variation it can control.
Hand-eye calibration
Establish the relationship between camera, robot base and tool correctly.
Select a 3D camera
Turn range, accuracy, material and cycle time into a shortlist you can test.
Next step
Test the actual part, container and gripper first
Photos and specifications provide direction, but representative trials reveal which grasps work, how cycle time varies and where intervention remains necessary.