Can AI solve the air traffic control problem?

In the coming weeks, college finals will be over and the unofficial start to the summer travel season will commence. If this year is like the last few, there could be concerns over airline and airport staffing levels, weather disruptions, flight capacity issues and, of course, canceled flights and delays.

But does it have to be that way? Or, more to the point, can artificial intelligence (AI) be used to prevent and mitigate these problems, as well as new one, the shortage of air traffic controllers?

To address the air traffic controller shortage, airlines are reducing their schedules in response to the Federal Aviation Administration’s (FAA) request to scale down airplane volume around the New York-Washington, D.C .air corridor. To maintain seat capacity may require flying larger airplanes, which could mean replacing regional jets with larger Boeing and Airbus airplanes on some routes.

Such a change would also require different types of flight crews (pilots are trained and certified on particular types of aircraft). This could absorb flight crew slack availability that the airlines hold for unexpected absentees or weather events during the summer, likely leading to more flight cancellations in the future. In the worst-case scenario, retraining flight crew to fly such equipment would be needed, which requires time — something that the airlines have little of in preparation for the upcoming summer travel surge. 

Air travel is a complex process. The absolute hard constraint is always safety (in contrast to schedule and service, which are more openly visible to travelers). And to ensure that airplanes and passengers remain safe, air traffic controllers are empowered to manage the air space and the ground movement of airplanes at and around airports. Yet, human error can occur, which means that there is always room for improvement.

Air traffic controllers essentially perform the same functions today that they have for decades. The radio communications and interactions with pilots are one-to-one. Such communications ensure that every pilot knows where their airplane should be, as they prepare for takeoff on the appropriate runways, to when they land at their destination and taxi to their arrival gate. The number of checks and balances to ensure safety within the air system is tremendous, and air traffic controllers are the lynchpin for such activities. 

Just as autonomous (self-driving) vehicles offer the potential for enhanced safety on highways, can artificial intelligence (AI) be injected into the tools employed by air traffic controllers to reduce their numbers while maintaining a commensurate high standard of safety? 

AI models are trained using large data sets that allow them to recognize patterns and support decision-making. There are concerns that AI will reduce the workforce, particularly for jobs that involve performing repetitive tasks.

Given the shortage of air traffic controllers, can AI be used to alleviate this shortage by taking on some of the most repetitive and perfunctory air traffic control tasks? Would AI effectively free air traffic controllers to focus their attention on their most critical responsibilities? 

Introducing AI into air traffic control cannot be done overnight. Nor should it be done expeditiously. It will require extensive testing and evaluation, across a multitude of simulated air traffic environments. However, if just 5 percent of air traffic controller activities can be assigned to and managed by AI systems, without safety being compromised, this could create an environment that requires fewer air traffic controllers for the same volume of flights.

AI has been criticized for taking away people’s jobs. Conversely, it can also be used to alleviate the impact of worker shortages in some fields. Perhaps air traffic control can be added to this list. 

When it comes to air traffic control, safety can never be compromised. This is why autonomous vehicles have been slow to enter our nation’s highways, with levels of autonomy added incrementally to vehicles. Nonetheless, autonomous features are becoming more ubiquitous. 

The same level of caution should be applied with transitioning some air traffic control tasks to AI systems. Research and testing will determine its feasibility. Yet, such efforts must begin now. Without such investments, the air traffic control system in 2030 may look the same as the one we have today. With even more volume and congestion in the air space, the same issues of concern will persist and grow even more critical.

Sheldon H. Jacobson, Ph.D., is a professor of computer science at the University of Illinois at Urbana-Champaign. He applies his expertise in data-driven risk-based decision-making to evaluate and inform public policy. He has studied aviation security for over 25 years, providing the technical foundations for risk-based security that informed the design of TSA PreCheck.