Why artificial intelligence is highly relevant to air traffic control

Skyway - Autumn/Winter 2019 banner

EUROCONTROL is at the forefront of introducing artificial intelligence into air traffic management, paving the way for new levels of predictability, safety and efficiency.

Air traffic management stands to benefit significantly from artificial intelligence (AI) by virtue of its reliance on repetitive activity – which lends itself to analysis and machine learning. In addition, much of the complexity is embedded in the driving factors that deliver safe air traffic control: for example flight planning, flow management, safety assessments and conflict prediction. It is no surprise the industry is adopting the technology to enhance both planning and operational activities, and early trials by EUROCONTROL reveal gains of between 20-30% in terms of predictability and efficiency.

AI has been an integral part of the European Commission’s Digital Single Market Strategy since 2017, supported by €1.5 billion co-funding under the Horizon 2020 programme from 2018 to 2020. This was followed by a roadmap and establishment of the European AI Alliance in 2018 to put Europe firmly on the path to becoming a leader in the AI revolution.

EUROCONTROL held an inaugural Forum on Aviation and AI in early 2019 which brought together key players and served as the launch point for a European AI Aviation Network. The activity is overseen by Directorate European CivilMilitary Aviation (DECMA) Director Philippe Merlo

“Our first role is to bring all the stakeholders together to enable an exchange of views. We want to define a roadmap with our stakeholders which defines our priorities and most important projects.” EUROCONTROL is a facilitator in this process, with no vested interests, Merlo says, “trusted for our neutrality and our expertise”.

The process is part of a wider activity underway which sets out to deliver an integrated solution coordinated by the Commission. EUROCONTROL plays a central role at a technical level within the Commission’s strategic initiative. Philippe Merlo points out the Agency already manages large quantities of ATM data in its role as Network Manager (NM) and this offers many opportunities for machine learning and AI. For example, the repository includes flight plans, radar tracks, meteorological data, safety incidents and special events, all of which can be interlinked to achieve a higher level of safety. With only a small percentage currently used, Merlo says the amount of relevant information extracted could increase dramatically with the help of AI: “We believe we can develop conflict prediction tools and complexity assessment tools which can mitigate risky situations much earlier in a flight.” The activity will extend beyond safety and capacity to address environmental factors and cost efficiency

An important first step towards achieving this level of connection is to complete implementation of the System Wide Information Management (SWIM) network as envisaged under Single European Sky (SES) legislation. This network interconnects all the ATM players – from airlines to airports and ANSPs – with seamless access and interchange of information between providers and users of ATM data and services. Simply sharing data in the first place brings huge benefit to operators in terms of better planning and collaborative decision making.

“Once you have the data, you are in a position to take the second step and apply AI to analyse this large volume of data and identify opportunities to increase efficiency,” says Philippe Merlo.

The first applications are air traffic control planning and flow management, where EUROCONTROL trials already show 30% improvement in trajectory prediction. Work to date looks at where safety can be improved, and delays reduced. Examples include the development of conflict and complexity prediction tools which can anticipate a risky situation based on past data, and can lead to mitigation much earlier in the flight.

“We already have tools which analyse trajectories automatically, can detect conflict risks and help air traffic controllers to avoid errors. Human errors are rare, thanks to controllers’ training and skills, however, they remain the biggest safety risk in ATC, and this technology can trigger an alert and present what-if scenarios.”

Philippe Merlo says environmental impact could be included in the future to expand the application to support optimised flight trajectories, and the agency is also investigating the airport environment where the technology could help to improve navigation around the airport surface.

Satellite navigation is another area where early research shows AI can add value. Satellite signals can be affected by atmospheric conditions which change with the seasonal equinox cycle. AI technology is being used to identify and isolate instances of ionosphere error to improve the prediction and likely impact of this phenomena.

Another promising area for the future digital environment is strengthening safeguards around cyber security. Philippe Merlo says AI can play a role in “monitoring the flow of data between your own ATM system and that of your neighbours. It becomes easier to identify abnormal data flows, or messages coming from unknown sources, to show where there might be a security breach.”

Within the framework of SES ATM Research (SESAR), EUROCONTROL is applying AI to support the latest re-categorisation wake vortex separations applied to ensure safe separation between aircraft on approach.

The Agency has amassed a large database from recategorisation research projects worldwide which is growing year on year. It includes trajectories from airports in Dubai, Singapore, Paris and London which are being used to determine the reliable safe vortex separation distance for different aircraft types.

The greater the volume of data, the better the outcome, especially when it comes to predicting flight profiles. EUROCONTROL currently exchanges flight data with the US Federal Aviation Administration (FAA) and service providers in the Middle East, and is in discussions with providers in Brazil, Japan, China and India.

Interested? Subscribe!

Never miss out on the latest developments from EUROCONTROL

This will lead to the exchange of flight-plan data from the moment a flight departs, resulting in more accurate trajectory prediction.

EUROCONTROL is applying the same principle within Europe to improve information exchange across all airports.

“Today the large airports are well connected, but almost a third of the traffic departs from smaller airports so we need this data too,” says Philippe Merlo. “We have started a campaign to encourage these airports to connect to the network.” Sharing data with Network Manager will lead to improved trajectory predictions.

Air traffic management is a human-centric activity and software development is viewed with caution by experienced operators. Experience has taught them to rely on their own skills in case of failure and the digital environment presents new challenges. EUROCONTROL is investing in training programmes for existing staff to establish trust in the new technology, and has launched a recruitment effort directed towards attracting skilled AI operators.

Longer term, EUROCONTROL is engaging with the European Aviation Safety Agency (EASA) to develop an appropriate certification process. Philippe Merlo describes this as another kind of challenge:

“When you work with AI, you are dealing with probability. It provides a prediction, not a determined outcome. Is it possible to certify a probabilistic tool? I believe it is possible but this involves working closely with EASA.”

The digital environment also needs to be secured against cyber-attack. This is especially important as data access grows. Data sharing needs to be secure and reliable to deliver quality information across the network. AI projects tackle these issues in preliminary project phases, for example by selecting only appropriate data and removing unwanted data before any analysis takes place.

“We are very much in favour of open data policy,” says Philippe Merlo. “We are developing specific AI tools which can be used to clean data before it is used. In a similar development, AI can also be used to monitor cyber activity by analysing abnormal data flows, however small.”

He believes AI will change the role of the controller for the better. “The main objective is to improve the efficiency of the working position. AI can give more priority to traffic management instead of tactically avoiding conflict. We’ve been working on this for 15 years, but thanks to AI we are now in a position to develop better advisory tools for the controller, and simplify the link between strategic and tactical air traffic management."

Read our full Skyway issue

Learn more about the new era for aviation powered by AI and our role in it.

Latest highlights

FLY AI #3 - AI in ATM: Enablers and use cases

FLY AI #2 - Research and Innovation in AI for aviation

FLY AI #1 - Partnering for AI in aviation