LE BOURGET, France — Will Roper, the U.S. Air Force’s acquisition executive, wants the service to shift to a faster, more modern approach for buying software and hardware. But that’s easier said than done.
The service has selected a number of programs where it will create technology in a new way, whether that involves creating digital prototypes, using agile software development or adopting a family-of-systems approach to address a changing threat environment.
To be successful, the Air Force will need to grapple with cultural and institutional pressures from those who champion tried-and-true acquisition practices. And in some cases, it also needs to prove to Congress that it is not taking inordinate risk and that its ambitions are executable.
Defense News spoke with Roper at the 2019 Paris Air Show on June 19 about some of the programs in which the Air Force is using new tools to develop technology.
Can you provide an update on Mad Hatter, the Air Force’s project to use agile software development to try to fix some of the problems with the F-35’s logistics system? You said in February that several improvements were to be fielded to the Autonomic Logistics Information System within weeks.
Mad Hatter is doing great. I have to give the team an A+ on being able to get started and start pulling apart the problems that our maintainers had.
They already deployed several apps that are helping maintainers. They fixed problems with the electronic equipment logs that were showing false positives, so those have been fixed, and the maintainers get to focus on things that are actually broken — not things that are reported as broken.
They fixed the scheduler, which had mismatches between the flight line system and ALIS, and they are currently working on things that are going to help maintainers do their own workflow on the flight line. There is a lot more to go for them. They’re putting Wi-Fi out on the line so that you can touch ALIS at the flight line, which currently you can’t. Maintainers have to go do their maintenance and then come back and enter data in the subsequent systems, and it doesn’t make sense to create data once and then replicate it again.
We want maintainers to be able to have ALIS in a protected, secure Wi-Fi network at the flight lines; that data is instantly uploaded. We’ve got work to go to get the accreditation done so that we could reach all the way back into the standard operating unit that touches Lockheed Martin. But we got a great partnership with Lockheed. They’ve been with us every step of the way.
What happens next?
I don’t have the answer yet, but one of the things that I think we should consider is the next variant of ALIS to be delivered. That’s 3.6. It’s currently going through negotiation and we’re approaching it as traditional ALIS, but if we believe in agile development, eventually we need to pull a development module of ALIS out of the traditional and put it into the Mad Hatter process. [Version] 3.6 is a candidate for that. If it’s not 3.6, is it 3.7 or 3.8?
The discussions we’re having now is about where’s the chalk line that we switch to the new methodology. We have to have enough development teams to do it and support the level and scope of the software, but I think we’re ready. We’ve got the team in the Air Force. We have 800 people in Kessel Run, [the Air Force’s software development team], that are currently doing amazing work for us.
With agile software development, you want to have exposure with the user. Once those apps were deployed, what was the feedback like? Did users want to see additional fixes, or were the apps coming out well already?
When final deployment was done, it was software as the users wanted. The users are involved from the beginning. Step one is the coders leaving their coding shop and going out to the flight line in Nellis [Air Force Base, Nevada], and sitting down, walking through how ALIS works and how the rest of the maintenance planning tools work. Understanding the pain points: What do you not like? What takes up your time? What do you want to change? Storyboarding that out to understand how it might be fixed, turning that into a development back log; so what am I going to attack and when? And then having the user touch products before they become final.
What the Mad Hatter team does is continues to iterate during design so that by the time you deploy, it’s in the image of what the operators have requested, not in the image of what the developers expected they wanted, and that’s the secret to “agile.”
Earlier this year, you said that once the Air Force hires an architect for the Advanced Battle Management System program, it can begin prototype work in the summer. Has that started?
We have. We’ve made great progress since we’ve hired our chief architect, Preston Dunlap. The Air Force has done a great job of rallying around him. ABMS and multidomain command and control is a top priority for our chief of staff, Gen. [Dave] Goldfein. He made it clear we’ve got to get ABMS over the goal line.
The rule that we're using at ABMS is we don't start talking platforms until the end. That's probably going to be biblical for us, for the first few years of the program. It is so easy to start talking about satellites and airplanes and forget what ABMS is going to have to uniquely champion. That's the data architecture that will connect them.
Preston’s done a great job of starting to write down requirements for our data architecture, leveraging a lot of work that’s already underway in our space-development portfolio. There’s an initiative no one’s heard of called Unified Data Library. It’s not a very exciting name, it’s like Castle Anthrax from Monty Python. But its super awesome. Unified Data Library is used currently for space situational awareness. Data pipes in through a variety of sources — commercial, academic, government — it’s able to be addressed in multiple layers of classification and we do microservices on top of it that are used by different users.
Step one, the thing we got to get right in the first year of ABMS, is that data architecture because people need to build systems that populate according to whatever standards we’ve fixed.
Step two, once we get the data architecture defined, will be the requirements for the population of the data. Maybe one sensor needs to be able to fill a gap that others are creating. We’re going to have to look at requirements at the system level and then tell satellites: “You need to be able to provide this level of data at this refresh rate. UAVs need to be able to [meet[ this rate" and so on and so forth.
Once we do that, then we'll be into the traditional part of the acquisition, which is building those satellites, building those UAVs.
What I’ve told Preston, we’ve got to do demonstrations along the way, so expect to do yearly demonstrations. The first one we want to get to is ad hoc mesh networking so that we get the same kind of internet of things effect where things start working together without humans having to control them and to not wait for the full architecture to be fielded. And it really will never be done; ABMS will merge over time as we put up satellites and UAVs. More things connecting to the network will make the system of systems better.
I think by the time we get to our 2021 budget, ABMS should be well definitized in terms of the lines of effort, the data architecture, you know we’ll have to have a line for artificial intelligence because we are not going to be able to pass all the data collected across the networks, a networking component, and then at the end, the platforms that provided. But the chief says it best: This has got to be about the highway, not the trucks that are on it. Step one is getting the highway paved.
How can industry accommodate this new paradigm? How do you monetize this to incentivize businesses?
That's a question I don't have all the answers to.
Openness in the internet of things makes sense because you can monetize the data, and that’s not going to exist for us. We’re going to have to have a contracting incentive that replicates it.
The best theory we have now is some kind of royalty scheme that the more open you are and the more adaptation we do at the top of your system, the more you benefit from it. If you create the system that allows us to put 100 apps on top of it, you benefit differently than if we can only put one. But the details are going to be difficult because maybe that one app is super important.
We’ve brought in industry early and we plan to have a series of industry days where we talk about what does it take to create that structure. Industry appears interested because if we can build open — yes, there has to be a way to get the profit and cash flow — but it gives more opportunities to bring new technologies in. But if we can’t replicate profit and cash flow on which their quarterlies depend, then they’re going to have to go back to the old model of saying they’re for open, but secretly giving you closed.
What are the immediate next steps for this program?
With the money that we have this year, we’re able to get this kind of data architecture, analysis and demonstration work done. The bigger money for ABMS really begins in ’20 and ’21. That’s where we’ll ramp up fuller-scale prototyping.
I’m actually glad that we don’t have big money this year because it can’t go build a drone or satellite. We’ve got to focus on the part that’s less sexy, which is that data architecture. We’re going to have to be able to do software development at multiple levels of classification and do it securely: All those are things that are hard to get people energized about but they’re going to be what makes or breaks for this program.
There’s been pushback from the House about the Air Force’s approach to replacing the B-52 engine. Lawmakers don’t like that you’re using Section 804 authorities. Why is that approach necessary?
It lets you get on contract a year and half earlier for an MDAP [major defense acquisition program]. And we found that for a variety of MDAPs, including the B-52 re-engine, you can use that year and a half to great effect. It’s much better to be on contract with industry and working from Day One than sitting around twiddling your thumbs.
In the case of B-52 re-engine, we are using that time to do digital engineering for the engine and pod integration, so we have all three industry bidders on one contract. We have Boeing on contract. They are working together as part of the source selection. They will deliver their virtual prototype to us by October. We would normally not even be on contract, and already we have a deliverable that will help us understand the challenges of integration. Are they able to keep the center of gravity and the fluid flow around the power pod? Are they able to keep that the same? We will have that earlier.
Because we have a year and a half, we can go from virtual prototyping into physical prototyping. We’ll select a vendor, and then we’ll do round two, which is: You gave us your digital twin, now give us the physical twin. Show us you can do it in the real world.
And then we’ll downselect and we’ll award a contract and move to the integration side of the program, move into the production side.
What I view in this program, because I had an extra year and half, I can spend more time in EMD, engineering manufacturing and development. I can take on more engineering rigor and retire risks faster than I would if I was denied that time and trying to meet a near-term operational need.
In the case of the B-52, the TF33 engine has been flown hard. It is an old engine. We have maintainers up in places like Minot [Air Force Base, North Dakota], that are doing heroes’ work to piece these engines back together. The depot is doing heroes’ work to try and do maintenance overhaul with a supply chain that is gone. We are at high risk for keeping that engine far into the future, and when my war fighter says I need a new engine and I have one path, which is the 804 path, which I can start a year and a half faster than I could the traditional path, I can’t tell that war fighter I’m going slow because that’s what others think is the best path. I can go fast and I can do it with rigor and discipline.
And to close off, why I don’t think it’s a wholesale overhaul of the acquisition system is that we do all of the same documentation we would have in a traditional program. We just do it after we award contract, so we get the benefit of jump-starting but we still do all the discipline and documentation. And I think the pushback we see from the Hill is a misunderstanding that we still do the rigor.