June 19, 2012 at 8:36 PM ET
Two researchers from USC published a study Monday describing a new kind of robotic sensor, modeled on the human finger, that can feel, explore and identify more than a hundred common materials. They hope the technology will be integrated into prosthetic hands and other devices that could benefit from a tactile sense.
The research grew out of a simple desire to investigate whether a robot could be made to identify textures; now the duo is starting a company called SynTouch, outside the university, to further develop the synthetic fingertip they created, with funding from DARPA, the NIH, the NSF and others. Clearly the work touched a nerve, so to speak.
Machines exist that can smell and taste, and of course see and hear, but touch has remained somewhat difficult to even define. Touch involves many sensations: pressure, temperature, vibration, and is as much behavioral as it is mechanical. In order to identify an object, for instance, you might naturally press harder or lighter, or move your fingers in a different direction.
So Professor Gerald Loeb and doctoral student Jeremy Fishel (who has since graduated) had to not only create a device that could replicate those sensations, but to act in a logical manner regarding how the material in question was to be explored. This led to both the BioTac sensor and what they are calling "Bayesian Exploration," (named after pioneering 18th century mathematician Thomas Bayes), a set of algorithms and rules that lets the robot explore objects intelligently.
The BioTac system is more or less an electronic fingertip: the outside is a soft and flexible material with an actual fingerprint pattern that helps in recognizing vibrations. Inside there is a fluid to give the fingertip a pliable but firm structure, and within that fluid, the sensors that detect pressure, temperature and sound (that is, vibrations).
It was trained on 117 materials, from wool to metal to paper, and can identify them correctly 95 percent of the time. It even performed better than humans overall, especially in distinguishing some highly similar samples.
The BioTac and its associated systems are still in development, and it will be some time before someone with a prosthetic arm will be able to tell the difference between table and coaster with their eyes closed. But the device must certainly be considered highly successful.
Devin Coldewey is acontributing writer for msnbc.com. His personal website iscoldewey.cc.