Land Warfare, Networks / Cyber

Can Army Intel Data Feed The Kill Chain?

WASHINGTON: On the future battlefield, “data is the holy grail,” says Brig. Gen. Robert Collins. But for the Army, like King Arthur and Indiana Jones, actually getting the grail is a daunting quest.

One of the biggest challenges for Collins, the Army’s new acquisition chief for tactical networks (aka PEO-C3T)? Getting data on potential targets from intelligence systems to combat units fast enough to strike them with the new long-range artillery and high-speed helicopters now in development. Digitally connecting the widest possible range of “sensors” to “shooters” in this way is the focus of Army Futures Command’s Project Convergence experiment starting this fall.

The types of data you want to access, Collins and other officers told a Potomac Officers’ Club webinar on Tuesday, range from full-motion video to electronic warfare detections of enemy transmitters. Quickly pooling that many kinds of data, from that many different sources, will require heavy use of artificial intelligence and cloud computing.

Cloud & Edge

“We’ve initiated two kinds of cloud pilots to help inform how we’re going to adopt cloud within the tactical environment,” Collins said. But, he cautioned, that’s very different from implementing cloud in the benign environment of an Army base in the US.

When people say data is “in the cloud,” that means it’s on someone else’s computer, usually a massive server farm somewhere, that you can access remotely. No access, no cloud, no data. But frontline units can’t drag fiber optic cable behind them wherever they go. Tactical networks have to move data over radio transmissions, which can be disrupted by terrain – mountains, reinforced concrete buildings, even dense forest – or by enemy hacking and jamming.

Project Convergence requires moving larger amounts of data over longer distances than ever before. “Our integrated tactical network is really being pushed to the limit,” said Maj. Gen. Peter Gallagher, the network modernization director at Army Futures Command, so extending its range and capacity are high priorities.

“We’re going to be extremely challenged by the bandwidth of the network,” said Leonel Garciga, information management director for the intelligence section (G-2) of the Army’s Pentagon staff. That means you can’t dump all the data on a single, central mega-server and expect everyone to access it over long-distance links. Instead, you need to selectively decentralize the data, pushing at least some crucial subset of it all the way out to the frontline edges of your network.

“Data that we don’t access routinely, we can host inside the cloud,” said Collins. “Data we need to access more rapidly and on demand, we can put into edge nodes.”

“We have to spend some time to really understand how we want to deliver data and to whom, when, and where,” Garciga said. “One of the things that we’ve been working on [is] expanding our Army commercial cloud provider services to edge nodes provided by the same vendor… How do you get that [data] in and out of a FOB [Forward Operating Base]? How do you get that on a plane? How do you move it around the world?”

“[There’s] a lot of emphasis on cloud, [but] we’re not trying to do all things in the cloud,” Gallagher said. “We’re trying to figure out what are the things we can containerize right away.”

Sorting all this data and moving it through the cloud requires automation. Today, moving data from intelligence collectors to combat operators is “a very labor-intensive process,” said Garciga. While the military could afford to throw people at the problem at the peaks of the wars in Afghanistan and Iraq, he said, in the future it needs to start “doing that work in a much more digital way that’s a lot more seamless and easier to do at scale and at a distance.”

Enter AI

One potential solution: artificial intelligence. This summer, Army artillery units in Europe did a live-fire experiment with targeting data provided by two complementary pieces of AI software, Prometheus and SHOT (Synchronized High-Optempo Targeting).

“Can I actually look at the [intelligence] community to help feed us tactically and provide us tactical data [for] deep fire targeting with long position fires?” said Alan Hansen, director of intelligence systems and processing at the Army’s C4ISR Center. “Prometheus is looking at all that [intelligence] imagery and being able to pluck out potential targets of interest, and doing it very successfully.” Prometheus’s output then feeds into SHOT, he said, “where we’re looking at how do we automate and use machine learning … to come up with the right information to address the attack guidance matrix so we can actually do effective calls for fire.”

“And then the last piece of this, which we will address probably on the follow-on phase is BDA, how do we do battle damage assessment?” he continued. “How do we use machine learning to determine that my effects actually were effective?”

Machine learning can even allow a machine to study the habits of successful intelligence analysts – like what queries they run in what databases for a given type of request – and implement those same behaviors as software algorithms, which can run much faster.

There’s no one omniscient algorithm that can solve all of these different problems. “There’s no one-size-fits-all machine learning capability out there,” Hansen said. “We’re on the verge right now of a big wave of machine learning capabilities coming out of industry…. Could I actually put multiple machine learning methods into my architectures?”

But no amount of new technology will be enough without the human element. Garciga and Hansen told the webinar that it will also take policy, procedural, and even cultural changes to streamline coordination between intelligence networks, which are governed by Title 50 of the US Code, and tactical ones, governed by Title 10.

“Policy and data governance, at scale …. whether it be at the enterprise level, or the at the tactical level – I think that that is a gap that continues to plague me, at least every day,” Garciga said. “[But] I think our bigger challenge is really going to be more of a cultural challenge.”

“Between the Title 10 and the Title 50,” Hansen said, “there is a lot of work to be done.”