Bin picking systems: what to prove before you buy

A bin-picking system must do more than recognize a loose part. It must choose an accessible target, grip it without creating a second problem, remove it from the container, and deliver it in a usable condition. Evaluate those steps together before committing to a robot, camera, or software package.

Describe the container as carefully as the part

Start the request for quotation with bin dimensions, wall shape, fill range, part quantity, allowable orientation, and replenishment method. Include photographs from normal production, not just a neatly arranged sample. Identify whether bins deform, arrive in several colors, contain liners, or sit differently after replacement.

Zivid bin-picking occlusion guide explains how bin walls and overlapping objects hide surfaces. A successful pick from the top center is therefore a weak acceptance test. Ask the supplier to address corners, parts leaning against walls, and the final items in a nearly empty bin. Define whether a residual quantity is acceptable and who removes it.

Describe the incoming parts by manufacturing state. An oily machined component and a clean display sample may demand different tests. If reflections are a concern, use the trial structure in our shiny-metal vision guide. Keep surface changes in the supported-product specification.

Separate perception from handling

MVTec 3D matching describes position and orientation data used in robot applications. In your trial log, separate “not found” from “found but unreachable” and “picked but not delivered.” Otherwise, improving a recognition threshold may hide a tooling problem or simply move failures downstream.

Ask whether an entangled part can pull neighboring parts out with it, whether thin parts can nest, and whether the chosen contact surfaces remain accessible. Compare candidate tooling using gripper finger design considerations. The camera cannot create mechanical access where the gripper has none.

Build a trial that resembles production

Prepare several independent bins rather than repeatedly resetting one favorable arrangement. Include the minimum and maximum fill levels, different container placements within tolerance, and representative process residue. Keep a subset of bins for a later acceptance run so adjustments are not judged only on scenes used during tuning.

Report accepted downstream parts, elapsed time, interventions, and residual parts. Record changeover and replenishment separately, then include them when estimating staffed output. Request both the average result and the range across bins. A proposal that sometimes needs lengthy manual recovery should not be evaluated only by its average successful cycle time.

Hypothetical example: a buyer tests ten bins and finds that the first half of each empties easily while the final few parts need rearranging. The next decision is not automatically a different camera. The team might change the bin, alter allowable residual quantity, revise the gripper, or conclude that structured trays are more economical. The test should expose that choice before purchase.

Specify recovery without relying on improvised intervention

OSHA industrial robot system safety guidance treats robot-system risk assessment and safeguarding as integration concerns. Have qualified personnel design intervention, replenishment, and restart procedures for the complete cell. Do not make an operator reaching into a running bin part of the normal production assumption.

The quotation should list retry limits, invalid-image handling, stuck-part response, and restart behavior after a stopped cycle. Clarify whether the robot resumes the same target or obtains a fresh observation. Require an operator interface that names the problem instead of reporting a generic failure code.

Buy a defined operating envelope

Finish with a written list of supported parts, bins, surface conditions, and delivery requirements. Include training for changing recipes, a backup procedure, and responsibility for future part qualification. Our vision-guided robot buyer guide can help decide whether random picking is justified at all. A constrained, well-proven application is often a more useful purchase than an impressive claim to handle anything.

When considering FAIRINO for bin picking, shortlist the robot only after drawing the deepest required pickup and the delivery path. The FAIRINO catalog gives you models to compare against that layout; the end-effector, selected camera system, and reachable grasp poses still need a joint feasibility review. Define what remains in the bin when the accepted cycle ends.

Request a FAIRINO bin-picking quote with bin dimensions, representative parts, gripper constraints, and delivery requirements. Ask the proposal to distinguish demonstrated conditions from items still requiring a trial.