DARPA's next big thing is … analog computing?
The key to mining those huge data sets collected from sensors and camera systems is to leave digital computing behind, the agency says.
The Defense Advanced Research Projects Agency (DARPA) has noted that the amount of data that various Defense Department agencies collect has increased exponentially in recent years, specifically the sensor data collected by intelligence, surveillance and reconnaissance (ISR) systems.
The increase in functionality, capability and sensitivity of the devices used to collect information, combined with the complex requirements of analyzing that data, means that current computer processing power may not be enough in the near future, DARPA says.
Since we are again pressing against the ceiling of the hard physical limits of current technology, the scientists at DARPA feel that another revolution is in order. That is why they have started a solicitation named the Unconventional Processing of Signals for Intelligent Data Exploitation (UPSIDE). In it they specify the need for a new type of processing based on the physics of microscopic (nano) devices.
The result would be, oddly enough, an analog computer, as opposed the digital kind you are probably using to read this.
A digital computer represents information through a series of bits that are in one of two states. This corresponds to a string of ones and zeros, which in turn represents the data. An analog computer uses some variable physical phenomenon, like electrical or fluid dynamics, to model the information.
DARPA says it’s looking to “break the status quo of digital processing with methods of video and imagery analysis based on the physics of nanoscale devices.” UPSIDE processing systems would work in a fundamentally different way from digital processors, with “physics-based” devices such as nanoscale oscillators capable of self-organizing and adapting to input, so they would not have to be programmed.
“Redefining the fundamental computation as inference could unlock processing speeds and power efficiency for visual data sets that are not currently possible,” Dan Hammerstrom, DARPA program manager, said in the agency’s announcement. “DARPA hopes that this type of technology will not only yield faster video and image analysis, but also lend itself to being scaled for increasingly smaller platforms.”
Digital processing usually serves our purposes just fine, which is good, because digital systems are inexpensive to make in mass quantities. But every so often a certain set of data is too large or its rate of accumulation is too fast for the capabilities of current digital computing technology. The last time this happened to DOD was during World War II, when they needed analog computers to process their gun data.
As government agencies at many levels -- from DOD to NASA to local law enforcement agencies -- collect ever-larger data sets from sensors, cameras and other devices, they may need a new kind of processing, leading up back to analog computing, albeit in advanced, nanoscale form. Everything old is new again.
Think you have the spark of an idea to get this new analog show on the road? Go download the solicitation for yourself. Good luck!