UK researchers at University of Exeter, Alan Turing Institute and NATS develop virtual airspace to train AI air traffic controllers
UK researchers have constructed a “digital twin” representation of airspace over England in an experiment to see whether artificial intelligence (AI) could aid human air traffic controllers or eventually replace them.
Project Bluebird, a partnership between National Air Traffic Services, the Alan Turing Institute and Exeter University, backed by public funding from UK Research and Innovation, presented its first results at the British Science Festival in Exeter this week.
Researchers argue AI could help find more fuel-efficient routes for aircraft, and may be able to cut delays and congestion to routes such as London Heathrow.
Such systems could potentially help deal with a shortage of human air traffic controllers, who take three years to train.
Richard Cannon, NATS research lead on Bluebird, told the Financial Times the agency had been preparing for such technology by recording air traffic movements over the UK for the past decade.
The data, including 10 million flight paths, is now being used to train Bluebird’s AI system.
Researchers have now brought together human air traffic controllers and AI systems to direct virtual aircraft through the “digital twin” simulation of the UK’s airspace.
Cannon said the project, which concludes in 2026, aims to run shadow trials in which AI agents operate on air traffic data in real time, so that their decisions can be compared to those of human controllers.
Such tests would not involve AI making real air-traffic routing decisions.
The project could lead to humans trialling the technology on an operational basis for several years, after which NATS and other air-traffic bodies could consider introducing an automated system.
The existing NATS computer system for processing flight data failed over the August bank holiday weekend after encountering a flight plan with contradictory data, resulting in the cancellation of more than 1,500 flights.
Exeter University professor of machine learning Richard Everson declined to comment on whether AI could have avoided that situation, but said such technology may be able to improve air traffic systems’ resilience and reduce the risk of failing when encountering unexpected events.
A study recently found 84 percent of businesses in the UK and Ireland have already begun using AI in some capacity, with 24 percent “confident” about the technology and having implemented it broadly across their organisations.