The JCF and the Combatant Commands: A Symbiotic Relationship

  • By: The JAIC
The JCF and the Combatant Commands: A Symbiotic Relationship Source: DVIDS

No Department of Defense component will have more use for—and be able to contribute more to—the Joint Common Foundation and its role in helping develop artificial intelligence capabilities than the Combatant Commands.

When it’s fully operational, the JCF will be a DoD-authorized, ready-to-use development platform with all the tools, frameworks, standards, processes, and other resources that DoD entities need to build, test, and field AI projects and applications.

While the JCF is essentially the plug-and-play foundation for AI collaboration, there are two critical resources it will need from its DoD partners to accelerate the development of AI capabilities and scale them out to the edge: data and use cases. As more strategic and tactical data is brought in and more use cases are built out by the JCF, other DoD entities will be able to leverage and build on that data and any previously built AI models and applications.

That’s where the COCOMs come in.

There are currently 11 COCOMs, thanks to the recent addition of the Space Command, and each has a unique geographic or functional mission that provides command and control of military forces in both peace and wartime.

The COCOMs and their subcommands are the controllers of the majority of warfighting-related and other strategic data, and they can also provide the broadest possible use cases—or the problem sets that AI can address—affecting the most users.

As such, the COCOMs will have a highly symbiotic relationship to the JCF, being both a prime contributor to and a prime recipient of JCF development and success.

Everything that’s done on the JCF “will touch a COCOM at some point,” said Colonel Bradley Boyd, lead for the Joint Warfighting Mission Initiative. The Joint Warfighting MI is one of six MIs, or AI projects, addressing high-priority and business reform challenges impacting all DoD and military organizations.

“Right now, we have these small development environments throughout the DoD, which are great, but they can’t handle the volume of data, the volume of development that needs to be done in order to scale AI across the force,” said Boyd. “That’s where the JCF becomes super important.”

As an example, Boyd explains, “CENTCOM (Central Command) might develop an AI algorithm to identify, let’s say, pickup trucks and then put all their full-motion video in one spot on the JCF. Then TACOM (Tactical Command) might be able to use that same algorithm and data, but tweak the algorithm a little bit and combine it with some of their data on other types of vehicles to meet their own unique needs.”

Not only does this collaboration accelerate each COCOM’s ability to develop and scale AI across their own components and subcomponents but also the Joint Force.

“You can’t do that type of collaboration, that type of leveraging of data and AI applications unless you have one unifying spot, which is what the JCF provides—a place where you can take all that data, store it, curate it, validate it, label it, and work together with other entities,” Boyd said. “And for this reason, the JCF is quite frankly the most important thing the Joint Artificial Intelligence Center will do for the next 5 to 10 years.”

From Theory to Practice

COCOM involvement in seeding and growing the DoD’s AI capabilities through the JCF is already underway.

As an example, the 160th Special Operations Aviation Regiment within the Special Operations Command is “partnering with the JCF to push their tools onto the JCF so that other forces can then leverage the AI models built by the SOCOM 160th SOAR,” said Lacey Duckworth, PhD, program manager for the Joint Logistics (formerly known as Predictive Maintenance) MI.

Joint Logistics is currently working to develop a repeatable, end-to-end AI ecosystem that will improve the DoD’s predictive maintenance at scale on a joint aircraft—the UH-60 “Black Hawk” helicopter.

One of the MI’s first products is the Engine Health Model, which was developed in partnership with Carnegie Mellon University. This AI model “predicts the probability of an engine hot start so decision-makers can consider next steps: Should I replace that engine or not? Or do I need to keep that aircraft back for training versus sending that aircraft out on a high-risk mission where, as predicted, it could possibly have a problem?” said Colonel Kenneth Kliethermes, Joint Logistics MI Lead.

That Engine Health Model has since been transitioned to the U.S. Army Special Operations Aviation Command’s 160th SOAR, which is using it to build additional tools.

One of their projects is the Model at the Edge Kit, which containerizes the Engine Health Model on a laptop computer and combines it with a download tool and a Web front-end display to increase the speed and efficiency of Engine Health Model recommendations.

“Now we have an all-in-one tool to download flight data from the aircraft, feed that data into the Engine Health Model, and view the predictions on a Web front-end in a matter of minutes” said CW4 Jason Slusser, Maintenance Technology Officer, Aviation Maintenance Support Office within USASOAC.

USASOAC personnel are currently testing the MaTE Kit in a controlled setting but will soon push it out to deployed environments for live testing.

Another issue that AI tools can help with is data accuracy, said Slusser. At present, maintenance workers manually input their work orders into a software program called Aircraft Notebook. “Oftentimes, they will enter an incomplete or incorrect work unit code which skews descriptive data analytics,” said Slusser. “In an effort to correct the “dirty data,” we are using natural language processing to analyze the fault narratives and apply a very specific work unit code that matches the text input by the maintainer.”

The Joint Warfighting MI is working with several COCOMs to build, test, and expand its Smart Sensor, a video processing AI prototype that rides on unmanned aerial vehicles and is trained to identify threats and immediately transmit the video of those threats back to manned computer stations for real-time analysis.

“You’re really looking at a dramatic reduction in the amount of data that has to be pushed back for a human to cull through,” said Boyd, noting that the Smart Sensor is being customized to work with unmanned aircraft used by the different services. “Instead of staring at one video feed and hours and hours of trees and rocks and nothing happening, that person can instead be monitoring ten video feeds because they are only seeing the stuff that really matters.”

The Joint Warfighting team will also be building an AI tool called the Fire Support Cognitive Assistant for the U.S. Marines. In time, Boyd expects his team will customize it for use within the Joint Force.

“You’ve got this tool that’s got a chatbot in it and some resource predictive analytics in it, and it will be great at the battalion level,” says Boyd. “But it can actually be applied at the CENTCOM to be a more tailored and strategic application, and that’s exactly what we hope to do in the future.”

The JCF: Where Sharing Translates to Growth

The JCF is critical to all this customization and expansion of AI capabilities, according to Duckworth. However, the ultimate key to AI success will be the COCOMs’ willingness to share data and products to the JCF so they can collaborate with other Joint Forces and leverage each other’s innovation and lessons learned. This is where USASOAC’s 160th Special Operations Aviation Regiment has taken the lead.

“They’re really eager to share what they’re doing, rather than close-holding that information,” Duckworth said. “And although their data is different and they do things differently because they’re on different program series, you only need to do a few tweaks or adjustments to their models and applications to make it work for other forces.”

Slusser notes this ability to share is the reason his organization has been working with the JCF from the start, and why they’re already pushing up various datasets, models, algorithms, and documented best practices for others to use.

“We understand that our aircraft are just a little bit different than what the Army, Air Force, and Navy have,” said Slusser. “But if we can get an 80 percent solution for the Army, for instance, and they can just take that and tweak it to make it work for their fleet, then they don’t have to start from scratch and vice versa.”

Even if COCOMs are just getting started with their AI projects, Duckworth encourages developers to work closely with the JCF and the different MIs because they can connect them with like-minded developers within other COCOMs. “What we have is a valuable opportunity here to learn, partner, replicate, customize, and extend each other’s work and breakthroughs,” Duckworth said. “If you can plug into what others are doing, you can save time and resources and put those savings into additional AI development areas critical to specific mission areas that are not currently being worked.”