Darpa: Fuse Nerves With Robot Limbs, Make Prosthetics Feel Real


Controlling robotic limbs with your brain is just step one. The Pentagon eventually wants artificial arms and legs to feel and perform just the same as naturally-grown ones. Which means step two is hooking up those prosthetics directly into severed nerves. That’ll allow the wearer to detect subtle sensations, respond to the brain’s neural signals, move with unprecedented agility, and “incorporat[e] the limb into the sense-of-self.”
Over the last decade, the Pentagon’s made remarkable progress in creating life-like prosthetic devices. And most of the advances are because of programs funded by Darpa, the far-out military research agency that’s also behind this latest project, called Reliable Peripheral Interfaces (RPI).

Already, Darpa’s funded ventures like the DEKA Arm, which relies on a joystick-style interface, and used “targeted muscle reinnervation surgery” for prosthetics that transmit neural signals from a bundle of nerves in the chest. Darpa-funded researchers at Johns Hopkins have even started human trials on their Modular Prosthetic Limb, which transmits cues to an artificial limb using brain-implanted micro-arrays.
But the RPI program taps into key shortcomings that persist in even the most sophisticated prosthetic devices. Existing neural-prosthetic interfaces aren’t sensitive enough to provide myriad signals — prototypes currently transmit around 500 events a second — or offer users a robust degree of freedom. Not to mention that current neural platforms have short life-spans and are tough to repair without invasive surgery, making them ill-suited to troops and vets in their 20s.
So Darpa’s after a prosthetic that can record motor-sensory signals right from peripheral nerves (those that are severed when a limb is lost) and then transmit responding feedback signals from the brain. That means an incredibly sensitive platform, “capable of detecting sufficiently strong motor-control signals and distinguishing them from sensory signals and other confounding signals,” in a region packed tightly with nerves. Once signals are detected, they’ll be decoded by algorithms and transmitted to the brain, where a user’s intended movements would be recoded and transmitted back to the prosthetic.
The end result would be a prosthetic that acts as a veritable extension of one’s own body. And a platform capable of accurately distinguishing between, and interpreting, different sensory signals — temperature, pressure, motion — would “allow the incorporation of the limb into the sense-of-self” and offer unprecedented freedom of movement for a prosthetic wearer.
The agency also wants an ultra-reliable platform, with an error rate of less than 0.1 percent and a lifespan of around 70 years. By comparison, current neural-recording interfaces last around two years before they need to be replaced. Sounds far-fetched, but Darpa’s already got one major lead: the agency’s new Neurophotonics Research Center will investigate fiber-optic prosthetic interfaces that can incorporate thousands of sensors into a single filament.