Researchers at the Massachusetts Institute of Technology and Princeton University have developed a "pick-and-place" robot with a standard industrial robotic arm equipped with a custom gripper and suction cup.
An "object-agnostic" algorithm enables the machine to evaluate a bin of random objects and ascertain how best to grip or suction onto an item amid the clutter, without knowing anything about the item before picking it up.
When the robot has grasped and lifted the object, cameras capture images of the item from multiple angles, and an image-matching algorithm lets the robot compare the images of the grasped object with other images to find the closest match, thus identifying the object before stowing it in another bin.
The team fed the perception system a library of product images from online sources, labeling each image with the correct identification while another learning algorithm related the pixels in a given image to the correct label for a given object.
From MIT News
View Full Article
Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA
No entries found