(A schematic showing the touch sensor against two different objects. The blue part of the images is the sensor and the yellow shows the layer with ridges on it, which simulates the function of human fingerprints.)
Robotic arms are commonplace, and robots can walk, see, hear and even smell. But giving them a sense of touch has been much more complicated. Now a team of scientists in Singapore thinks they have part of the answer: give robots fingerprints.
Fingerprints on people come from the tiny ridges on the fingertips, and they do things like increase friction (so you can hold a glass) and tell us the way an object is textured and curved. Robots have to rely on visual cues. But that isn't as efficient because a robot would have to "memorize" a huge amount of data for that to work. On top of that there are a lot of things people do using only the sense of touch, such as buttoning a shirt.
So, Saba Salehi, John-John Cabibihan and Shuzhi Sam Ge of the National University of Singapore came up with the idea of putting ridges on a sensor. The ridges would transmit the force from an object in a certain way, and help a robot understand its curvature. Human finger ridges do something similar — when you touch something an object touches different parts of your fingers, so a ball feels different than a cube.
They built two sensor arrays just four millimeters on a side. One had a latex layer with ridges on it and the other was smooth. The researchers also programmed the sensor with a machine-learning algorithm. By alternately pressing on the sensor with a flat and curved shape, they found the one with the ridges on it provided better feedback than the smooth one did. The computer was able to tell the difference between the shapes more than 97 percent of the time.
Robots that can work with a touch sense would be able to do a number of things without needing elaborate visual recognition systems — for example, a real touch sense could tell the machine that it was holding a ball and direct it to grasp differently. There might also be applications in prostheses, teaching computers to simulate the feeling of round-ness or square-ness.
Via Sensors (open access)
Image: Sensors 2011, 11, 8626-8642