Sign In 
Sign In
Sign In
Robotics
Community of Robotics Enthusiasts: Discuss latest robots and questions with fellow robot lovers
605 Members
Join this group to post and comment.
Algorithm That Improves A Robot's Grasping Skills
Now get ready for a robot whose experience on earlier situations shall guide it on what to do in new situations. Cornell Researchers have come up with a new algorithm which, in their own words, lets the robot acquire skills from its past experience.
![[IMG]](proxy.php?image=http%3A%2F%2Fwww.crazyengineers.com%2Fwp-content%2Fuploads%2F2012%2F05%2Frobot1.png&hash=8b2a42e91152ed45ede0271cba9427d4)
source: publications.csail.mit.edu
Lipson’s gripper comprises a flexible bag loaded with granular material. Employing the new algorithm, the robot gets a 3D image of the object to analyze a series of rectangles that complement the size of the gripper and examines every rectangle individually for several features. When the robot encounters a new object, it selects the rectangle having the highest score as per the rules found. The shape and size of the object plays a key role for the robot to get a stable grasping point.
The test involved connecting a Microsoft Kinect 3D camera and a jamming gripper to the robot arm. Attempting to pick up 23 objects, the robot's success rate over 90%.
source: publications.csail.mit.edu
Hod Lipson, associate professor of mechanical engineering and computer science, along with Ashutosh Saxena, assistant professor of computer science and a specialist in machine learning, developed the "hardware agnostic", and they claim that it'll function on any kind of robot gripper. They took inspiration from the "universal jamming gripper", and their paper is slated for presentation at the International Conference on Robotics and Automation in St Paul, Minnesota, coming 16th May. Yun Jiang and John Amend have co-authored this paper.
Lipson’s gripper comprises a flexible bag loaded with granular material. Employing the new algorithm, the robot gets a 3D image of the object to analyze a series of rectangles that complement the size of the gripper and examines every rectangle individually for several features. When the robot encounters a new object, it selects the rectangle having the highest score as per the rules found. The shape and size of the object plays a key role for the robot to get a stable grasping point.
The test involved connecting a Microsoft Kinect 3D camera and a jamming gripper to the robot arm. Attempting to pick up 23 objects, the robot's success rate over 90%.