Waltham, Massachusetts 2451
We’ve been commercializing the technology for use in industrial inspection, 3D printing and forensics. However, robotics companies have recently been contacting us with interest in using our devices as part of robotic grippers. Initial experiments with compact and inexpensive cameras and lighting have produced interesting results.
Bill Yost, Co-founder and CEO
Janos Rohaly, Co-founder and CTO
Micah Kimo Johnson, Co-founder and Chief Scientist
Edward Adelson, Co-founder
# of Employees
Problem Your Company/Products Solve
GelSight’s technology allows robotic grippers to do three things that would otherwise be difficult or impossible: assess the orientation of a grasped object, actively prevent slip and analyze surface features/texture of objects.
Since GelSight can sense surface features while in the grasp of a robot, those features can be compared against a template to determine the degree to which an object has shifted or rotated from its expected orientation. An example is in picking up and inserting a USB cable. Work with Robert Platt’s group at Northeastern demonstrated that a Baxter robot could be taught to insert USB cables using GelSight.
Slip occurs when frictional forces are insufficient to keep an object in a grasp. Friction is manifest as a tangential force making it difficult to measure using conventional contact sensors (which typically only measure normal forces). Since GelSight sensors can measure both tangential and normal forces they can be used as part of a closed-loop control protocol for grasping objects. Our videos show a Baxter arm picking up a variety of objects using just enough force to hold them without slipping and not so much force as to break them. The gripping parameters were not changed between objects.
Finally, GelSight is able to turn surface texture into 3D information in the form of a height map. These data can be used to recognize objects or identify defects in much the same way that human fingers do. In fact, GelSight sensors give human-like capabilities to robotic hands, but at a resolution that exceeds human sensitivity.
Ping Fu, Omega Cambridge SPV