Chips By Lee Mathews Sep. 19, 2014 5:17 pm
It’s not always easy being a robot. While it’s painfully simple to perform certain tasks, others that appear simple can prove to be frustratingly difficult. Take plugging a dangling USB cable into an iPhone charger, for example, with a two-pronged pincher.
That used to be impossible for Rethink Robotics’ Baxter, but he’s mastered it now. Baxter recently received a major dexterity upgrade: a brand new fingertip sensor that lets him position his pincher with tremendous precision, even when a subtly swaying object is the target.
The glow you see is from a GelSight sensor, a 2.5D imaging sensor that was created by MIT researchers back in 2011. GelSight combines a gelatinous slab with a metallic skin with a series of cameras that surround it to record imagery. Its creators refer to it as a scanner, since it’s able to capture information about an object’s surface and store it digitally. The first-generation GelSight was sensitive enough to recognize the raised ink patterns on a $20 bill; newer versions are accurate to depths of less than a micrometer.
The GelSight sensor that Rethink put into Baxter’s “fingertip” isn’t quite that precise: a bit of sensitivity was lost while scaling it down to the required size. It’s made a huge difference to Baxter, though, who can now deftly grab onto a cord and plug it effortlessly into a port . Northeastern professor Robert Platt says that’s because he now has “enough information to localize the pose of the object that [he's] holding.”
This could greatly improve job satisfaction among robots; they’ll no longer be limited to picking and placing objects that have a fixed position. On the downside, that means one with the right AI onboard could figure out how to improvise a garotte…
Now read: MIT’s Cheetah robot gets let off the leash to run and jump outdoors
Photo credit: Melanie Gonick/MIT
More...
That used to be impossible for Rethink Robotics’ Baxter, but he’s mastered it now. Baxter recently received a major dexterity upgrade: a brand new fingertip sensor that lets him position his pincher with tremendous precision, even when a subtly swaying object is the target.
The glow you see is from a GelSight sensor, a 2.5D imaging sensor that was created by MIT researchers back in 2011. GelSight combines a gelatinous slab with a metallic skin with a series of cameras that surround it to record imagery. Its creators refer to it as a scanner, since it’s able to capture information about an object’s surface and store it digitally. The first-generation GelSight was sensitive enough to recognize the raised ink patterns on a $20 bill; newer versions are accurate to depths of less than a micrometer.
The GelSight sensor that Rethink put into Baxter’s “fingertip” isn’t quite that precise: a bit of sensitivity was lost while scaling it down to the required size. It’s made a huge difference to Baxter, though, who can now deftly grab onto a cord and plug it effortlessly into a port . Northeastern professor Robert Platt says that’s because he now has “enough information to localize the pose of the object that [he's] holding.”
This could greatly improve job satisfaction among robots; they’ll no longer be limited to picking and placing objects that have a fixed position. On the downside, that means one with the right AI onboard could figure out how to improvise a garotte…
Now read: MIT’s Cheetah robot gets let off the leash to run and jump outdoors
Photo credit: Melanie Gonick/MIT
More...
