Bin Picking

A robot identifying and picking parts or items, one at a time, out of a cluttered bin.

This is a classic task in industry and logistics: parts or goods are piled randomly in a bin, in a mix of orientations, occluding or even tangled with each other, and the robot must first use a camera, usually 3D vision, to identify a target, estimate its pose or a graspable point, then plan a path that avoids the bin's walls, and pick items out one at a time with a suction cup or gripper. In manufacturing this is often called “3D vision-guided bin picking,” used for machine tending and feeding parts onto a line; in logistics it means picking a single item out of a tote. Amazon's Picking Challenge, held from 2015 to 2017, turned this into a hot topic in robotics research. The main remaining difficulties today are transparent, reflective, and soft objects, along with entirely new items the system has never seen.

ExampleAt an auto-parts factory, a structured-light camera photographs a bin of jumbled metal parts, an algorithm computes each part's 6D pose, and the robot arm picks them out one by one and places them at the machine's loading position.

Related

Grasping · Grasp Pose Detection · 6D Object Pose Estimation · 3D Vision-Guided Robotics · Machine Tending · Amazon Picking Challenge

Sources

Bin picking – Wikipedia