3D vision robot bin picking for randomly presented parts.
Automate the selection of individual components from bins, stillages or totes using 3D vision, application-specific gripping and collision-aware robot motion.
Bin picking is a perception, gripping and recovery problem.
A successful cell must continue to make valid picks as the pile changes, not only demonstrate one favourable part position.
Identify usable part geometry despite variation in depth, orientation and pile state.
Choose pick points the tool can reach without striking neighbouring parts or the bin.
Orient parts for a machine, fixture, conveyor, inspection or kitting process.
Respond predictably to occlusion, double picks, failed grasps and an empty or unpickable bin.
Validate the difficult states before committing to the cell.
Random parts create more uncertainty than conveyor-based pick and place. Feasibility work should focus on the least favourable pile conditions and the production consequences of a missed pick.
- Test representative bins at full, half-full and near-empty states.
- Check shiny, dark, transparent, oily or low-feature surfaces under production lighting.
- Define whether parts can be separated, shaken, recirculated or re-presented after a failed scan.
- Model robot reach into corners and the collision volume of the full end effector.
- Confirm whether the destination needs a fixed orientation, inspection or secondary datum location.
- Plan bin exchange, empty-bin removal and uninterrupted production where required.
The minimum bin-picking system is more than a camera and robot.
A credible design combines the sensor, calibration method, part model or learned detection, grasp rules, collision model, robot path planning, gripper feedback, destination logic and a defined response to unsuccessful picks.
- 3D sensor selected for working distance and surface behaviour
- Rigid calibration between camera, robot and cell coordinates
- Grasp candidates filtered for accessibility and tool clearance
- Collision model covering bin walls, parts, tooling and surroundings
- Grip confirmation before the robot leaves the bin
- Operational dashboard for pick success, retries and recovery events
What evidence should be supplied for a bin-picking assessment?
Physical parts, the real container and an understanding of part variation allow the vision, tool and motion concept to be tested against realistic conditions.
- At least a representative quantity of good production parts.
- Bin, tote or stillage internal dimensions and maximum fill level.
- Part CAD where available, including variants and known dimensional changes.
- Required successful picks per minute and acceptable retry behaviour.
- Destination fixture, machine interface or required placement orientation.
- Information on oil, swarf, dust, surface reflections, entanglement or part damage.
Frequently asked questions
These answers support initial planning. Final performance, safety and scope are confirmed against the actual product, process, environment and acceptance criteria.
Is 3D vision always necessary for bin picking?
It is normally required for random three-dimensional piles. Parts presented in a controlled layer or known plane may be handled more simply with 2D vision or mechanical presentation.
Can the system pick tangled or interlocked parts?
Some entanglement can be managed through grasp strategy and re-presentation, but strongly interlocking parts may need mechanical separation or a different container strategy.
What happens when no valid pick is visible?
The system can re-scan, try an alternative grasp, move the pile, request operator intervention or change bins according to the agreed recovery sequence.
Can a cobot be used for bin picking?
Potentially, where payload, reach, speed and the complete risk assessment support it. Random-bin motion, sharp parts and tooling can still require guarding or other protective measures.
Continue planning the project.
Use these closely related pages to define the task, compare the technical options and prepare a stronger automation brief.
Robot vision systems
Select 2D or 3D sensing, calibration and inspection architecture.
Explore robot vision →Robot grippers and EOAT
Develop a tool that can access reliable grasp points.
Explore gripping →Robotic pick and place
Compare random-bin picking with controlled conveyor presentation.
Explore pick and place →Discuss the production task with an automation engineer.
Send product details, target output, available space and a photo or short video of the current process. We will identify the next technical step.