acm-header
Sign In

Communications of the ACM

ACM TechNews

Helping Robots Get a Grip


robotic hand simulation software

Credit: Columbia University Robotics Lab

Columbia University robotics researchers Peter Allan and Matei Ciocarlie have developed a new way to control a dexterous robotic hand. The researchers realized that while human hands have 20 joints that can each bend, each joint is not capable of moving completely independently, and each joint's movement is connected to the movement of other joints. Existing software attempts to account for all the degrees of freedom in a robotic hand's joints, but this is computationally difficult and slows down the actions of the robot. Instead, Allen and Ciocarlie aimed to limit the hand's joints in the same way a human's hands are limited to allow them to control a complicated robotic hand with faster, more efficient algorithms while maintaining full functionality.

The researchers experimented with four robotic hands, each with multiple joints, and developed software to control each gripper by linking its joints. During testing, the software was able to quickly calculate grasping positions to pick up different objects. The system works by determining the angle at which the hand is approaching the object and then choosing the hand position that will provide the most stable grasp. Then, if a human controller believes the position is correct, a command is given to take hold of the object.

"Grasping objects with a human-like hand is a seemingly complex computational problem," says Georgia Institute of Technology professor Charlie Kemp. "This work . . . shows that a complex hand may not require a complex brain."

View a video entitled "Hand Posture Subspaces for Dexterous Robotic Grasping."

From Technology Review
View Full Article

 

Abstracts Copyright © 2009 Information Inc., Bethesda, Maryland, USA


 

No entries found