Big Data: The US Air Force's air and space operations centers collect, analyze and share massive amounts of data around the clock. (US Air Force)
WASHINGTON — For the past decade, technological improvements have made the dream of the unblinking eye a near reality. Cameras are cheaper to make and install, while providing higher quality images. The advent of streaming video, much of it relayed via satellite, has given US Defense Department analysts a wealth of information previously unimaginable.
The unintended consequence of those increased capabilities is that the Pentagon is drowning in data, unable to keep up with the gigabytes of information collected every day — and it is only going to get worse as more capabilities come online.
It is a large enough problem that the National Security Agency included a special subject line for research on “coping with information overload” in the National Intelligence Program Summary, also known as the “Black Budget.”
The agency requested $39 million in fiscal 2011 and $64.3 million in 2012 for that research. That request dropped to $48.6 million for its proposed 2013 funding, according to documents disclosed in August by the Washington Post.
The US Air Force, which runs a significant portion of the Pentagon’s space and airborne ISR assets, also is aware that its operators and analysts are in danger of being swamped by a wave of information.
One approach that could offer relief is the automation of data processing.
“For me, the bigger bandwidth [challenge] is bandwidth from the system to the human, and how we really optimize that more effectively,” Mica Endsley, the Air Force’s chief scientist, said in a July interview. “Right now, we have people sitting there watching each one. That gets to be cost prohibitive. So there is research going on about automatic processing of imagery.
“Lots of it is long periods of boredom where nothing happens,” she said. “Can we have automation help with that, and key in the human when something interesting is going on [that] we want eyes on? I think there is room for improvement there, and there are things we can do with effective integrations between automation and people to be able to interpret that data more efficiently.”
Executives from the Virginia-based Exelis agree with Endsley that automating ISR functions is going to be key.
“We need to start processing [the data] without an overabundance of human intervention,” said Rob Mitrevski, vice president for Environmental Intelligence/Integrated Sensing and Information Solutions at Exelis. “What we really want as humans is the answer. And the closer we can get to the answer without having to tell each and every piece uniquely what to do, the better off we’re going to be.”
Ideally, a system could be set up that looks something like this: An ISR asset is tasked to track a house and is told to note whenever someone enters the front door. If that criterion is met, it sets off another chained automated task — for instance, telling the asset to request another platform with a hyperspectral camera to come and check for a specific substance whenever someone enters the house.
In theory, it would operate like a classic “if-then” statement, used by computer programmers for years.
Meanwhile, operators on the ground are no longer receiving gigabytes of data indicating that nothing is happening. Instead, they are being sent only relevant data to work with, leaving them less mentally taxed and potentially trimming the number of operators and analysts needed for each mission.
Mitrevski is quick to point out that he doesn’t foresee a “Terminator” situation, in which an algorithm can be set that leads to a military strike.
“There is a real emotional issue with a detect-to-kill chain because it allows the machine to make that decision,” Mitrevski said. “That’s not our intent. Our intent is to get the person making that decision all the information they need without delay.”
Chaining those assets together in an automated fashion poses strategic and technical challenges.
For instance, what happens if two satellites request the same Predator UAV for separate missions? How would different mission commanders share their assets? Those are questions the Pentagon would need to work through before automation could become a reality.
Disaggregated Space Architecture
The technical challenges should be easier. When asked how long it would take for an automated system to be set up if given an unlimited budget, Mitrevski estimated around a year, much of which would be focused on creating algorithms.
In addition, the potential move toward a disaggregated architecture for space — replacing the small number of exquisite space systems with a larger network of smaller, cheaper assets — offers an opportunity to upgrade onboard space-based processing systems at a faster rate than before.
To handle the level of automated analysis being discussed, ISR assets require an increase in onboard processing. A few years ago, that would be hard to imagine. Increasing the processing power on these assets would require upgrades in power and weight — two things at a premium on space systems.
In a white paper released in August, Air Force Space Command noted that a disaggregated architecture would provide more opportunities for the latest technology to reach space-based assets.
Current space systems have a developmental lifespan of up to 14 years, and often last a decade or longer once launched. While it is possible to update software, the capabilities of the satellites are mostly locked in place once they are launched.
Moving to a disaggregated architecture, with multiple units going up over time, would mean that satellites can have much more relevant technology during their lifespan.
And with the continual growth of processing power, that could give an ISR asset all of the thinking power it needs.