How to do autonomous flight right

It seems like military-grade autonomous flight technologies are being announced every other week. But that breakneck pace should actually cause concern. Simply put, too many of these touted solutions prioritize announcing demos in a sandbox over nailing the true fundamentals. If we don’t build and scale autonomous flight the correct way, American warfighters and our national security could pay a severe price.

To their credit, the Defense Department is putting out crystal-clear demand signals: autonomy is the future of air power. Last year’s Exercise Mobility Guardian 23 proved that maximizing endurance and optimizing pilots’ time will be essential in Indo-Pacific conflicts. Air Force Special Operations Command is working to use more unmanned aircraft under its Adaptive Airborne Enterprise concept, and not a year after the announcement of the widely discussed collaborative combat aircraft program, major awards are already being awarded. Air Force Secretary Frank Kendall has been an ardent advocate for autonomous flight: “It’s a security risk not to have it. At this point, we have to have it.”

Autonomy has seen a surge of investment in recent years, with hundreds of millions of dollars pouring into the development of new technologies. As a result, a myriad of cutting-edge solutions have flooded the market. But as someone who has spent over six years laser-focused on this challenge—my company Merlin is working to create a non-human pilot that’s able to fly a wide variety of aircraft—I’m concerned that too many companies are in a cycle of splashy marketing moments rather than electing to adhere to two critical tenets.

First, many of the technologies rolling out today are far too narrow for practical use. They ignore the basic aviate, navigate, and communicate functionality needed to safely execute a mission. Take remotely piloted solutions. Their effectiveness still hinges on a human-operator link, which can be severed by electronic warfare, communication malfunctions, or even bad weather. In contested or degraded environments, these limitations present enormous operational risks, and not just to the mission itself. True functional autonomy requires systems that don’t need a human in the loop and can seamlessly perform all aviation, navigation, and communication functions independently.

But functional autonomy solutions alone are not enough. To create a truly effective autonomous package, they must be coupled with tactical skills: dogfighting, swarming, EW, etc. In an ideal world, the customer would integrate functional and tactical autonomy together, but this is incredibly difficult to do—maybe impossible. The interdependency of both stacks is simply too great for channeled software, and customers are better served with a full-stack “pilot” that can handle both functional and tactical autonomy.

Second, new autonomy capabilities must advance incrementally and collaboratively with the warfighter to build full trust. Autonomy in aviation is still an emerging technology . Solutions cannot be developed in isolation by tech companies and delivered pre-baked to the military. It may not be flashy, but you must crawl before walking, and walk before flying. Companies must spend thousands of hours working directly with operators to understand where their product roadmaps need to lead not just months, but years, into the future. Even reducing crews from two pilots to one requires hundreds of factors to make flight operationally viable and comfortable for that remaining pilot. Companies skipping these stages are wasting resources and endangering lives—neither of which will ensure ongoing investment.

I’ve seen too many startups show up with the bravado of having “solved” autonomy, but when you peel back the layers, they’ve glossed over or completely ignored these basic requirements. It’s akin to a self-driving car company claiming their vehicle can perfectly parallel park itself, but can’t actually accelerate, brake, or turn properly.

While I mentioned the government has put out clear demand signals, they could also be more helpful in reinforcing the need for companies to meet these requirements. With autonomy being such a priority, there’s an eagerness to find that “future-state” solution that can leapfrog today’s limitations. But that’s dangerous when dealing with the profound complexities of airborne systems and human lives. No competent auto company would let the allure of delivering a flying car distract them to the point of missing today’s vehicle- safety standards.

Our military needs continued aggressive development of autonomous flight capabilities, and rapid development is largely the domain of the tech industry. Adversaries like China aren’t standing still. But that advancement must happen through a tight cycle of government-industry partnership, rigorous testing of proposed solutions against proven systems engineering principles , and demanding that any autonomy upgrades are thoroughly validated at each incremental step by pilots themselves before moving to the next stage.

Autonomous flight represents a potential force multiplier that will tilt the balance in future conflicts. But that competitive advantage disappears if we suffer catastrophic incidents due to reckless rushes toward autonomy while disregarding fundamental airmanship and Human Factors. We must stay grounded in reality, focused on the grueling stepping stones, and earn the complete trust of our aircrews. Cutting corners on autonomous flight is truly an unacceptable risk.

Matt George is the founder and chief executive officer of Merlin, an aerospace startup that is building a non-human pilot to enable reduced-crew and uncrewed flight.