Bulk bin to machine fixture
A robot removes parts from a tote and loads a machining, inspection or assembly fixture. A regrip station can establish the final datum before loading.
Random-bin robotic feeding
3D bin picking systems for parts supplied in bulk containers.
Locate components in a tote or bin with 3D vision, select an accessible grip point and transfer each part to an inspection, orientation or machine-loading station.
- 3D depth vision
- Robot-guided gripping
- Tote and bin handling
- Flexible part presentation
Direct answer
What is 3D bin picking?
3D bin picking is a robotic parts-feeding method in which a depth camera measures randomly piled components in a container, software estimates the position and orientation of accessible parts, and a robot selects a suitable grip before transferring the component to a known location. It can reduce dedicated feeder tooling for suitable applications, but success depends on part visibility, grip access, bin geometry, robot reach, cycle time and a reliable strategy for difficult or depleted bins.
Technology overview
Bin picking replaces some mechanical orientation with sensing and robotic decision-making.
Conventional feeders progressively separate and orient parts through mechanical tooling. Bin picking instead observes the bulk pile and chooses a part that can be gripped. This can be attractive for larger components, multiple variants, low-to-medium volume production or parts that are awkward to run through a bowl. It can also remove a manual transfer step between storage totes and an automated station.
The technology is not automatically suitable for every random pile. Highly reflective or very dark surfaces can make depth sensing more difficult; thin overlapping components may hide usable edges; cables, springs and flexible items can interlock; and deep bins can restrict the robot wrist or gripper. The project must assess the least favourable bin condition rather than only a shallow, well-spread layer of parts.
Many cells use a secondary station after the initial pick. The robot may place the part onto a regrip fixture, turntable, camera station or flexible feeder so final orientation, quality checks and downstream presentation can be completed consistently. Cycle-time modelling should include failed picks, image acquisition, collision checking, regrip and bin-change time—not only the robot movement for one ideal part.
Where it fits
Where 3D bin picking can add value.
The strongest applications combine accessible component geometry with a clear business reason to avoid manual unloading or dedicated feeder tooling.
Several part variants in one cell
Recipes, gripper change and model data can support a controlled family of parts where separate dedicated feeders would occupy more space or complicate changeover.
Parts unsuitable for small vibratory feeders
Castings, mouldings and robust fabricated parts may be more practical to pick from a container than to orient through conventional bowl tooling.
Feasibility decisions
A successful bin-picking cell balances vision, grip and pile behaviour.
The part must be detectable, reachable and removable without collisions, while the cell continues to recover as the bin empties or the pile changes.
3D visibility
The camera needs enough surface information to identify accessible parts. Material finish, colour, reflectivity, thin edges and depth shadows should be assessed with real components and containers.
Gripping method
Vacuum, parallel fingers, magnetic grippers or custom tooling must tolerate pose variation and lift the selected part without disturbing the pile or trapping adjacent components.
Collision management
Software and cell geometry must protect the camera, wrist, gripper and bin walls. Reach at the corners and bottom of the container should be checked in simulation and physical trials.
Pile depletion
Pick availability normally changes as the bin empties. The cell may need bin agitation, a tilt mechanism, a shallow tote, a tray insert or an operator/bin-change request before parts become inaccessible.
Final orientation
The first grip may not match the downstream pose. Regrip fixtures, 2D vision, turntables or a second robot move can create a stable datum before loading.
Recovery and throughput
The cycle model should include no-pick results, re-imaging, grip confirmation, dropped-part recovery, container exchange and any manual intervention required at low fill levels.
Project definition
Information needed for a 3D bin-picking feasibility review.
Provide parts and containers together. A CAD model alone does not show how components settle, overlap or become hidden in a real pile.
| Component samples | Representative production parts, accepted variants, weight, dimensions, surface finish, flexibility and damage constraints. |
|---|---|
| CAD and datum | 3D model where available, critical features, acceptable grip areas and the final downstream datum or orientation. |
| Container | Tote or bin dimensions, wall shape, maximum fill depth, liner, component quantity and planned change method. |
| Target cycle | Normal and peak parts per minute, machine cycle, required buffer and permitted recovery time. |
| Grip constraints | Surfaces that may be contacted, vacuum suitability, magnetic properties, finger access and permitted clamping force. |
| Vision environment | Reflectivity, colour range, contamination, ambient light, camera mounting limits and whether the bin can be stationary during imaging. |
| Downstream interface | Fixture, conveyor, nest, machine door, inspection station or regrip requirement, plus confirmation signals. |
| Cell constraints | Available footprint and height, robot preference, guarding, operator access, bin logistics and utilities. |
Feasibility before commitment
Test the deepest, most depleted and most occluded bin conditions.
A convincing trial should begin with a full production-style container and continue until the remaining components are difficult to access. Record successful picks, re-images, no-pick events, grip failures, collisions avoided and parts left behind. This gives a realistic view of effective output and operator involvement across the whole bin, not just the easiest first layer.
Common questions
3D bin picking questions.
What parts are suitable for 3D bin picking?
Rigid components with detectable surfaces, accessible grip areas and manageable overlap are often the best candidates. Very thin, transparent, highly entangled or deformable parts can require pre-separation or another feeding method.
Is bin picking faster than a bowl feeder?
Not necessarily. A dedicated bowl feeder can provide a very high and continuous output for a stable component. Bin picking is chosen for flexibility, larger parts or reduced dedicated tooling; the complete effective cycle should be compared.
Can one cell pick several different part types?
Yes, within the validated camera, gripper, robot and software capability. Each variant still needs model setup, grip definition, collision review and physical trials.
Does the robot place the part in its final orientation?
Sometimes. Where the first accessible grip is unsuitable for final loading, the robot can use a regrip nest, turntable or secondary vision station to establish the required pose.
What happens when only a few parts remain in the bin?
The cell may continue picking, request agitation, tilt the container, ask for a bin change or leave an agreed residual quantity. The depletion strategy should be defined in the acceptance specification.
Related pages