Cornell University researchers have developed a algorithm that enables a robot to learn grasping skills from experience and apply them to new situations.
The gripper consists of a flexible bag filled with granular material. As the bag settles on an object it deforms to fit around it, then air is sucked out of the bag, causing the granules to pull together and tighten the grip.
Since modeling how a deformable bag shapes around irregular objects is too difficult to compute, the researchers used a learning approach. In a three-dimensional object image, the robot studies a series of rectangles that match the size of the gripper and tests each one on a variety of features. The robot is trained on images of many objects until it has built an archive of features common to good grasping rectangles. When presented with a new object, it selects the rectangle with the highest score according to the rules it has determined. The robot also weighs the size and shape of the object to choose a stable grasping point.
In a test in which it had to pick up 23 objects, the robot succeeded an average of 90 percent to 100 percent of the time.
From Cornell Chronicle
View Full Article
Abstracts Copyright © 2012 Information Inc. , Bethesda, Maryland, USA