3D Bin Picking: Why Depth Is the Technology That Makes It Work

Bin picking is not a new problem. Manufacturers have wanted to automate it for as long as robot arms have existed. The challenge is that parts in a bin do not cooperate. They stack on top of each other, lean against the bin walls, sit at random angles, and look entirely different depending on which face happens to be pointing up.

Early attempts at robotic bin picking either required upstream fixturing that defeated the purpose, or used 2D cameras that produced flat images without any information about depth. Neither approach worked reliably in unstructured bins. The robot consistently missed parts, collided with the bin, or picked parts at angles that caused downstream handling failures.

3D bin picking changed that. By adding depth to the camera's view, the system can see the bin the way a human hand-eye system does: in three dimensions, with full awareness of which part is on top, how it is tilted, and exactly how far away it sits. That spatial data is what makes reliable automated bin picking possible.

What 3D Adds That 2D Cannot

The fundamental limitation of 2D bin picking is that a flat image is ambiguous about depth. A part that appears at the center of a 2D image could be sitting on top of a pile or buried underneath others. A part that looks horizontal in the image might be tilted 30 degrees toward the camera. A grasp point that looks correct in 2D might be pointing at empty space in 3D.

A 3D vision system eliminates that ambiguity. It produces a point cloud: a spatial map of the bin contents where every visible surface has an X, Y, and Z coordinate. The vision software analyzes this map to determine which parts are accessible, which are on top, how each part is oriented in three-dimensional space, and what approach angle gives the robot a clean grasp path.

Three specific capabilities define good 3D bin picking performance.

Recognition on difficult surfaces- Metal parts are reflective. Dark rubber absorbs light. Transparent plastic parts scatter it. Structured light cameras handle all of these by projecting a known pattern and measuring its deformation rather than relying on ambient lighting conditions. Mech-Mind's Mech-Eye industrial cameras produce accurate point clouds on dark, reflective, and complex surfaces that defeat most standard depth sensors.

Deep bin access- As a bin empties, the remaining parts drop lower and the arm must descend further to reach them. The 3D vision system tracks this in real time, updating pick coordinates to match the actual depth of the remaining parts rather than assuming a fixed reference level. This is what allows a bin picking cell to run to empty without requiring manual intervention when the level drops.

Collision-free path planning- With a full 3D map of the bin, the path planning software can route the arm around the bin walls, camera brackets, and neighboring parts on the way to the target grasp point. It selects approach angles that avoid collisions and retract paths that do not disturb parts left in the bin.

The AI Layer

Modern 3D bin picking systems add a machine learning layer on top of the geometric 3D data. This is what enables recognition of parts that vary in appearance, castings with parting lines, components that look different from different angles, or mixed bins containing multiple part types.

Mech-Mind's approach uses AI-powered algorithms alongside the 3D point cloud to identify the best pickable candidate in the current scene. The system evaluates each visible part for accessibility, stability of grasp, and approach clearance, then ranks candidates and selects the optimal pick. When the first-choice pick fails, the system falls back to the next candidate automatically without stopping for operator input.

The combination of geometric 3D data and AI-powered candidate selection is what produces the pick success rates that make 3D bin picking practical in production rather than just in demonstration.

Which Applications Need 3D Bin Picking

3D bin picking is not the right tool for every application. For parts that always arrive in a consistent, fixed position, simple 2D vision or even a fixed-program robot is faster and cheaper. 3D bin picking earns its cost in applications where variability is unavoidable.

Metal machined parts from CNC operations arrive in bins in random orientations with reflective surfaces. 3D structured light handles the surface conditions; depth data handles the orientation variability.

Automotive components including fasteners, brackets, and sub-assemblies that feed assembly stations from bulk bins require 3D guidance to pick reliably at the required cycle rates.

E-commerce piece picking from storage totes containing multiple SKUs requires the vision system to identify a specific item type among others and pick it without disturbing neighboring items.

Food processing for irregularly shaped items including proteins, produce, and packaged goods where no two items are identical in shape or position.

Which Arms Handle 3D Bin Picking

The arm requirements for 3D bin picking center on reach, payload, and 6-axis wrist flexibility.

For light-to-medium bin picking with parts under 5 kg, the Fairino FR5 ($6,999) is the strongest starting point. Its 924 mm reach, 6-axis flexibility, and full ROS compatibility make it straightforward to integrate with 3D vision platforms including Mech-Mind's Mech-Vision and Mech-Viz.

For heavier parts or bins where the gripper weight plus part weight exceeds 5 kg, the Fairino FR10 ($10,199) provides 10 kg of payload capacity with the reach and 6-axis maneuverability needed for deep bin access and complex approach angles.

For demanding applications where heavier components or deep bins push payload and reach requirements further, the Fairino FR16 ($11,699) adds payload headroom while maintaining the 6-axis flexibility that 3D bin picking requires.

Getting Started

Use our Cobot Selector to match an arm to your bin picking requirements, or the Automation Analysis Tool to model the ROI against your current manual sorting process. Browse our full Fairino lineup and UFactory cobots with current pricing, or book a live demo to see a 3D bin picking cell in action.

FAQ

Why does bin picking require 3D vision?

Bin picking requires 3D vision because parts in a bin are stacked in three dimensions at random orientations. A 2D camera cannot determine depth, which means it cannot tell which part is on top, how steeply it is tilted, or what approach angle the robot needs for a clean grasp. 3D spatial data eliminates that ambiguity.

What makes a 3D camera good for bin picking?

The most important qualities are accurate point cloud generation on difficult surfaces (reflective metal, dark rubber, transparent plastic), sufficient depth range to cover the full bin depth from full to empty, and fast enough scan rates to keep up with the robot's cycle time. Industrial structured light cameras are the standard choice for demanding bin picking applications.

Can a cobot do 3D bin picking?

Yes. Cobots with 6-axis flexibility, sufficient reach and payload, and open API integration support 3D bin picking effectively. The Fairino FR5, FR10, and FR16 all meet these requirements and integrate with industrial 3D vision platforms through ROS and open communication protocols.