Opinion: Machine Learning Is Coming To Your Flight Operation

Private aviation has a running joke that the industry is at least a decade behind the rest of the world in its technology adoption. Grease boards still deck the walls of some flight departments, but these days, most have been replaced with flight management systems, engine monitoring programs, and more modern ways of tracking the myriad pieces of data that flow through a flight operation.

Everything in aviation is complex, from the regulatory environment, the extensive supply chain, uncontrollable weather, highly automated aircraft, and beyond. For software developers in business aviation, it’s quickly discovered that every operation is unique, and it’s often challenging to provide a “one size fits all” technology solution in an industry with such complexity.

For artificial intelligence developers, aviation could be an incredible opportunity to save the industry substantial costs and reduce accidents.

“Today in civil aviation, technology can only improve any of the variables–safety, cost, or capacity–at the expense of at least one of the others. To move the envelope on all three simultaneously requires a disruption towards new technologies,” says John Mora, director of communications at Daedalean. “Technologies such as Daedalean’s AI-enabled systems exemplify these new technologies.”

Alaskan Airlines has already demonstrated the benefits of AI in aviation. In 2021, the carrier conducted a trial of Flyways AI by Airspace Intelligence to optimize route planning. The six-month trial resulted in significant fuel savings, conserving almost 500,000 gal. of Jet A by providing more efficient routing for nearly a third of the carrier’s planned flights.  

“The system autonomously evaluates the operational safety, ATC compliance, and efficiency of an airline’s planned and active flights,” the company says. “When it finds a better route around turbulence or a more efficient route, it provides actionable recommendations to flight dispatchers.”

This is the mission AI was designed for: to process hundreds or thousands of variables and augment human decision-making. Scheduled airlines are complex, but unscheduled air carriers are even more so in terms of planning and adapting. AI is here to assist, not replace, human intelligence.

Daedalean, a Swiss AI program developer, is at the forefront of transforming the way pilots fly. Their technology, leveraging data from cameras and Machine Learning, provides pilots with “Situational Intelligence.” It acts as a digital co-pilot, enhancing safety by alerting crews to threats, which will one day include uncharted wires, wildlife, and other hazards. It can even provide vertical landing guidance for helicopters with a centimeter level of accuracy.

Credit: Daedalen

Today, their “PilotEye” cameras and other components are attached to the aircraft. Other tools are in development, such as ground collision avoidance, emergency landing guidance, and full integration of onboard systems like radar.

It’s expected to be the first machine-learning-based safety-critical application certified in civil aviation.

Garmin has been rolling out its Autoland feature on more airframes, most recently the King Air 200 series. Its “Autonomí” technologies continue to expand from electronic stability and protection to emergency descent mode, rudder bias correction, and a smart glide feature during engine-out emergencies.

When asked if we need to be worried about the impacts of AI “taking over” our planes or our jobs, Mora emphasizes: “Daedalean uses neural networks, which are, in a conventional classification, a subset of machine learning, which is, in turn, a subset, in these terms, of “traditional” AI. The applications are deterministic: given the same input, they always give the same output, just like classical software. They are designed to perform a limited set of tasks, not to evolve toward reasoning capabilities.”

Alternatively, Generative AI technologies, like ChatGPT, create new information and make decisions based on what it learns. The future applications of Generative AI in aviation are wide open, but some have suggested using it to scour through maintenance records and identify premature part failures.

Fly-by-wire technology has been used in business jets for 17 years, boosting the range and passenger comfort of the aircraft with these systems, while airliners began using it 36 years ago. The adoption rate has been slower due to a need to understand the benefits and the sheer cost and scale of designing these systems.

On the other hand, AI tools could be one of the first technologies to blossom from general aviation before commercial airlines. Rapid advances in machine learning could provide the industry with inexpensive ways to enhance pilot decision-making, like aftermarket cameras that could identify deer on a runway at night well before a human eye could see it.

The most significant barriers will be the length of time it takes to certify these systems and the effort needed to gain the trust of the pilots who use them.

Jessie Naor is the author of the Sky Strategy column in BCA and is CEO of FlyVizor, an aviation M&A advisory and business consulting firm. She is a former founder and president of GrandView Aviation.