Col. Charles Wells (U.S. Army)
Col. Charles Wells is the program manager for the Army’s Distributed Common Ground System, the service’s effort to integrate all streams of intelligence.
There has been controversy in the past, driven by a link-analysis software company, Palantir, which was reportedly favored by some Army units. Capitol Hill even weighed in on Palantir’s behalf.
That fire has died down, and C4ISR Journal editor Aram Roston interviewed Wells about the DCGS-A, the challenges of bandwidth, and the way DCGS-A is now working with Palantir.
What’s the central message about DCGS-A?
Everything from link-analysis tools in Afghanistan to signal-intelligence tools in the Pacific, that’s all part of DCGS-A. Any sensor out there, space satellites, aircraft, terrestrial sensors, even the soldier who might be on a checkpoint collecting biometrics data, coming on to a [forward operating base] in Afghanistan, all that is put into the same system.
It’s the system that brings all the pieces of the puzzle together. You can connect all the dots in a single system. It’s powerful.
And then the second reason is the collaboration. So, the fact that the Navy has a DCGS program, the Air Force has a DCGS program, we’re building to a common architecture. As I build the product in DCGS Army, I could send it over to the Navy and say, “Hey, if you guys got any more intel on this,” they can say, “Yeah, I have some intel. Let me add something to that and pass it back to you” or the same thing with the Air Force. So you kind of get this collaboration with the DCGS programs within the joint community. That’s been very powerful. So, it’s the ability to see the whole picture and then to collaborate and share intelligence in the joint community and with intel community partners that’s made DCGS really what it is today.
So key to that has been an open architecture. We, consciously, within the DCGS community, have been building to what we call this DI2E — Defense Intelligence Information Enterprise — and it’s really a common architecture where all the DCGS programs build to open standards, they publish the standards. We say, “Here’s how we’re meeting the standards,” and then we share data. The technical way we do that is called the DCGS integrated backbone.
You know, that’s fascinating. The collaboration doesn’t mean that you guys kind of ingest any things from the other services, right?
We do. Yeah, absolutely. We’ve got over 600 data sources, some of them are from the other services, some of them are from intel community partners, [National Security Agency], [Defense Intelligence Agency]. We ingest all that data, over 600 data sources.
I mean you get U-2 feeds and all this stuff?
That’s what comes in to you guys, as well as the Air Force? That’s kind of new because that was the problem, wasn’t it? Integrating intel?
Well, yeah, historically — that’s it exactly. That’s been the historic challenges every service had, its own stovepiped intel. And then within the service, you had the sensor stovepipe down to their own processing system.
So, that’s really what DCGS has done. It’s kind of broken those barriers and allowed everybody to collaborate more.
So it’s not automatically ingested but you can ask for it or — right?
So those 600 data sources are coming in all the time. Then, as I generate a product, I want to see what the Air Force had, that’s where I used the DCGS integrated backbone and say, “OK, what does the Air Force have on these topics?” and I can look over and try to share and pull products from them and share with them.
But you’re not automatically getting that in the first place?
No, you have to do a query through the DIB.
In the Air Force, they’re doing their own, what we call PED — processing, exploitation and dissemination — so they might have their own sensor called Blue Devil, which I saw in Afghanistan. But they bring down the Blue Devil data and they do the PED ... and then they have a product there, and then as I am on my DCGS Army system, I might say, “Hey, I want to take a look at some Blue Devil results.”
And then they’ll say, “OK, here’s what our Blue Devil got over yesterday.”
The Navy, I guess, is the same?
Same thing with the Navy.
What are the obstacles you still have to overcome in terms of integration?
One of the obstacles is bandwidth, particularly in Afghanistan. And this has been a challenge on the FOBs. So you got so many sensors out there and one of these forward operating bases, you’re bringing a lot of data down to that FOB. The challenge is if there’s some really robust data there and I want to bring it all the way back, not just in here but back to, like you said, an Air Force location somewhere in the world, how do you do that if you got very small bandwidth?
Given very small bandwidth and all the data that’s now at the end of the architecture, how do you get that and bring it back and share with everybody. And that’s where we look to industry. One of our big pieces of the programs is industry partnerships. So industry might have compression technology.
You mentioned encryption. They got new encryption technologies. There’s a product that we’re looking at to use limited bandwidth but to move large data files over a limited bandwidth, things like that.
Let me ask about the analysis part. You guys have actual DCGS analysts?
Yeah, they’re DCGS analysts. It’s a military occupation specialty for the Army called 35Fox. Trained at Fort Huachuca. They are what we call the all source analyst, so they specify in all different types of analysis using the DCGS-A system and they are the subject matter experts.
But we’ve got other military specialties for specific areas of intel that also use DCGS.
So you might have a single intelligence specialist in the Army, you might have an inventory intelligence, human intelligence, but they all use the DCGS.
How do you download all this in the field? Because downlinking everything would use massive bandwidth, right?
What’s interesting is that you acquire the footprint at the brigade because you have to preload the data.
So if you deploy, you would take the [combatant command] set of data, preload that data into your server suite and, then this way, you’re not as reliant on the comms any longer.
Oh I see.
So let’s say you’re going to the Horn of Africa and you know you’re not going to have a lot of communications there.
We could configure what we call a tactical edge node. You probably want to find out about the hydrology for your patrols and things like that. We’ll load all that up on a server, the tactical edge node, and you take that as you go to Horn of Africa.
So now, even though you don’t have connectivity back to continental United States, you still have all the data you need to conduct your mission.
OK, you got all that data, but now you’ve got hundreds of streams, Predators, and whatever is flying over you. How do you access that?
So we directly download that into a component DCGS-A. Again, that is the tactical ground station.
So as a brigade commander, you get your own tactical ground station so you can directly downlink all those UAV feeds to your forward operating base and then you can store it in your tactical edge node.
The bandwidth seems to be staggering.
I think the best way is to give an example. When we talk about this cloud node, this is what we call a cloud core node. We actually deployed one to Bagram, Afghanistan, and we sent that, it stood up in April of 2011. We think it was the first tactical deployment of a cloud computing node in the Department of Defense. When we say tactical deployment, we push a cloud node into a combat zone. We got it up and operational.
That’s 1.4 petabytes, so pretty robust storage.
Are you moving to a place where you can actually get software to replace some of these?
We have a partnership with industry and with some government labs on things like that.
We have a partnership with [the Defense Advanced Research Projects Agency] now.
It’s called the Insight program and they’re looking in at things like that. One of the things that they’re looking at are algorithms to comb through video and to point things out for you as an analyst.
So for example, traditionally, it has been very manually intensive to go through full-motion video, you got hours and hours of it.
One of the things that we’ve been partnering with DARPA is I’m looking for [improvised explosive device] emplacers, and the algorithm is somebody digging or somebody is carrying something specific to a road, the system can go through and you can say, “Hey, in eight minutes and 32 seconds, take a look at this because I found somebody digging.” So the analyst can immediately go to that spot in the video.
Who does that?
DARPA has been doing that. The overall program is called Insight VIRAT. It is one of the tools, which was the video analysis and exploitation tool and they’ve got other tools under Insight that were partnering what they’re trying to bring into the program.
What I consider drudgery is watching eight hours of full-motion video feeds to look for the IED, the guy placing an IED. What we can do is we can tell the analyst, you know, here’s six different spots on that video where we think it’s happening.
So what we can do is make it more efficient for the analyst, but you’re never going to actually replace the analyst, because let’s face it, the most important tool for intelligence analysis is the soldier.
You’re not going to have a magic box.
Right. There is no silver bullet that’s going to spit out the intelligence to answer, but these are tools to help the analyst do a better job.
Maybe too science fiction. Are there yet systems that can help go through full-motion video, detecting people who have weapons?
Sure. Yeah, you can go through full-motion video footage and you can say if someone has a weapon, point that out to me.
The software exists to do that?
Right, and it’ll go through, just like we said, someone in placing an IED.
It can also point out people with weapons. That’s going to depend on the resolution of the video.
Let me ask you about Palantir.
Right. We’ve officially — when I say we, I mean the big Army — the Army and, specifically, Research and Development Command, has signed this cooperative research and development agreement, this CRADA, where Palantir is now in the lab up here at Aberdeen [Proving Ground, Md.] and we’re looking at some of their key technologies. So some of their key technologies, one of them is ease of use.
So if you look at Palantir, it’s easy to learn, easy to use, and specifically down at the lowest text level. So you’re down at the company level. Maybe you’re not a trained intel analyst but you want to be able to use the system very quickly, Palantir has been pretty capable in that ease-of-use perspective.
The second technology we’re looking at with Palantir is called Nexus Peering, which is the ability to synchronize computing nodes even though we don’t have the bandwidth.