Getting a Grip
A new approach lets robots grasp objects more quickly.
Source: “Hand Posture Subspaces for Dexterous Robotic Grasping”
Matei T. Ciocarlie and Peter K. Allen
The International Journal of Robotics Research 28: 851-867
Results: Researchers at Columbia University have developed control algorithms for robotic hands–motorized grippers with three, four, or five fingers–that reduce the number of calculations required for the devices to grasp an object. Using the algorithms, a robotic hand was able to grasp a wide variety of objects, including a wine glass, a telephone, and an ashtray.
Why it matters: The advance could make robotic hands for prosthetics more useful by enabling them to grasp objects more quickly. The hand could be attached to a prosthetic arm, which the user would maneuver into place. As the hand neared an object, the system would position the fingers for the best possible grip. The person would then only have to push a button to trigger the hand to close around the object.
Methods: Conventional control strategies for robotic hands independently calculate the position each joint must assume to grasp an object. The researchers eliminated some of these calculations by using algorithms to virtually link the movements of the joints, so that the angle of one joint determines the angle of others. To test the algorithms, the researchers made three-dimensional scans of selected objects, which helped the system calculate which finger positions would produce a stable grasp.
Next steps: Whereas the current system must be programmed with the shape of an object, the researchers plan to develop a sensor system that allows the hand to grasp new objects on the fly.