Magic Finger: Robot Digit Can Tell What It Touches
|
|
Researchers created this robotic finger, which can stroke a material and identify what it is.
CREDIT: From "Robots Get a Feel for the World" by USC Viterbi on Vimeo |
Two researchers have created a slightly squishy, life-size robotic finger that can identify materials by touch. The unique technology in the finger might someday give prosthetic hands a sense of touch, or help people judge the textures of consumer products in factories.
"I'm excited because it's a new class of device," Gerald Loeb, a physician who teaches biomedical engineering at the University of Southern California and one of the finger's creators, said in a USC interview.
The finger, named BioTac, has a hard core of sensors covered first in a layer of incompressible liquid and then a layer of rubbery, elastic material. The elastic skin has tiny ridges stamped on it, similar to a fingerprint, which help transmit vibrations to the sensors in the device's core. The liquid-and-skin layering makes BioTac a bit soft, but mostly firm, so it "mimics human finger pads," according to a paper Loeb and his colleague published yesterday (June 18) in the journal Frontiers in Neurobiotics.
The sensors inside the finger are able to detect force, vibrations and temperature. (The paper's researchers studied only vibrations.) Human fingers also feel for textures in part by interpreting how the skin vibrates against a surface when the finger is dragged along it. Large vibrations mean something rough, while no vibrations means something smooth, such as glass. [9 Cyborg Enhancements Available Right Now ]
Loeb and his colleague Jeremy Fishel, an engineer, gathered squares of 117 materials such as paper, glass and carpet, which they bought in art supply, fabric and hardware stores. They rigged their robotic finger to a lever controlled by a computer program they wrote. They then had the program drag the finger over each material, applying different forces and speeds. The computer program "learned" each sample, creating a 117-texture database in its memory.
Afterward, they tested the program, giving it one of the 117 textures at random. The program used what it learned during the training phase to choose at what speed and pressure to touch a sample to test it. After an average of five finger-drags per texture, the program correctly identified samples 95 percent of the time. When asked to distinguish between two similar textures that human volunteers couldn't tell apart, the program was 99 percent accurate.
Discerning textures is an important part of the normal functioning of hands, so the robotic finger's abilities may make it especially helpful in prosthetics. "If you can't feel what you're touching, it slips, you can't identify things," Loeb said. "By adding tactile sensing to prosthetic hands, we can overcome that problem."
