Is AI cleared for take-off?19 December 2022
As digital air traffic control towers continue to gain acceptance across the industry, the implementation of AI in this space is beginning to gain traction. Andrew Tunnicliffe speaks with Andy Taylor, chief solutions officer, digital towers, at NATS, to learn about how machine learning and AI can be paired with digital towers to improve operational efficiency, safety and resilience, and help to maximise situational awareness for air traffic controllers.
“Digital tower technology tears up a blueprint that’s remained largely unchanged for 100 years,” said NATS’ operations director, Juliet Kennedy, at the end of April 2021. She was proclaiming the arrival of a remote digital air traffic control (ATC) tower for London City Airport (LCY). To be clear, it’s not the first airport in the world to make use of this technology – that honour belonged to the Scandinavian Mountains Airport in Dalarna, Sweden, in 2019. LCY was, however, the first major international airport to follow suit. Nestled at the heart of the city’s Docklands, the airport serviced 5.1 million passengers in 2019; a figure that fell to 714,000 as aviation suffered in the darkest days of the pandemic. The news followed a decade of research and development, live trials – including at the airport during the Covid lockdowns – and even operational services on a smaller scale.
ATC is complex and all-consuming, where multiple elements are in play: from looking out of a window to tracking flights via radar, monitoring ground movements to surface winds and other fluctuating weather conditions, and countless other functions. Bringing this together, in one place, unquestionably has its advantages.
“These are complex areas to survey, monitor and basically maintain that scan; particularly as humans were designed to look in one direction,” says Andy Taylor, chief solutions officer at NATS. Digitally having that information in one view is enhancing the role of the ATC.
The system at LCY comprises 16 high-definition (HD) cameras and sensors on a mast to capture a 360º view of the airfield, fed through independent secure fibre networks to 14 HD displays providing a panoramic image to controllers at NATS’ Swanwick control room. They have live visual and audio footage, supported by radar information overlaid with call signs, altitudes and speeds of aircraft approaching and leaving; weather readings; and the ability to track other moving objects.
As has been shown on a large scale in Norway, this allows operational control of multiple airfields from one site. In June, Avinor, the state-owned civil airports operator, opened the world’s largest single remote digital tower in Bodø. The facility would become home to 15 remote towers covering the equivalent number of airports. It was a major milestone for Avinor, after opening its first remote tower at Røst Airport in Norway in 2019.
AI and data on the same path
This digitisation, however, creates significant amounts of data – arguably the next frontier for aviation. Earlier this year, while speaking to International Airport Review, Taylor said that less than 10% of data produced in Europe was being used, making it an untapped resource needing to be unleashed. He was speaking specifically on how the towers receive data can benefit operations on the ground – fuelling, aircraft spacing and so on. It’s a view he holds still. “Along with all the other operational data, we record everything from a digital tower. It’s normal that we supply a 30-day archive, that’s usually the regulatory requirement. But in some places we’ve deployed systems that have up to 90 days; then we store data externally, for an infinite time. So, you’ve got all of that additional data, available in real time, with a millisecond latency and the ability to harvest those archives.”
This doesn’t just cover ATC – it provides an overview of the entire airfield and all the events that are happening. It’s also helping advance another huge development too – machine learning and artificial intelligence (AI). Taylor believes this is the most exciting prospect in improving efficiency, safety and resilience at the world’s airports today, particularly when combined with the digital and remote tower concept. To understand what role AI might play, NATS and the Alan Turing Institute are working on a five-year project, supported by a £3m government grant through from the UK’s Engineering and Physical Sciences Research Council. Project Bluebird aims to develop the world’s first AI system to control a section of airspace, using digital twinning and machine learning technologies for the ‘safe and trustworthy use of AI’. Combining these elements could herald the next generation of ATC.
Each operation, input and assistance that AI provides can help inform what it might do in the future – essentially layering knowledge to continually build a response rationale. This ability is being capitalised on while training the technology, Taylor explains. For example, using archived data, information could be pulled from the previous winter to train the AI model in what it might potentially face in the future. He is, however, keen to stress this is not happening operationally. “That, certainly in our safety-critical environment, isn’t an approach we would take. We always deploy trained AI models that are closed,” he notes. “So, at this stage, they’re not able to continue to learn when they’re in operation.”
AI training is something Project Bluebird is focused on, aiming to develop a probabilistic digital twin of UK airspace as a real-time, physics-based computer model that predicts future flight trajectories and their likelihoods, “essential information for decision making” the Alan Turing Institute says.
A helping, digital, hand
The project also hopes to bring greater understanding of how AI and machine learning can be fused with human interactions to increase efficiency and safety. Taylor says there is already evidence that bringing digital capabilities, joined with the expert decision making of controllers and AI, to ATC increases the effectiveness of operations. “When it comes to either a human, or particularly an AI, making decisions, the more information we have available [for] them, the better the decisions,” he says, adding that in these scenarios accuracy of the outcomes is improved by 14%, according to NATS analysis.
As anyone familiar with the digital tower concept knows – providing a full view of the airfield and airspace above it, overlaid with data that ATCs would ordinarily have to look away to retrieve, is a huge step for aforementioned efficiency and safety. Bringing AI and machine learning into that realm could go even further – continually assessing the environment to identify risks the human eye might miss.
“AI can be trained to monitor the full 360º for specific things that we need the controller to be aware of,” Taylor says. “That means controllers don’t have to be looking in that direction as part of their scan – the system can flag that they need to now look somewhere else, that an important or safety-critical decision or monitoring is required.”
The benefit of this data-driven, digitised capability stretch far beyond just seeing aircraft land and take off – it reaches across the entire turnaround process. It may also advance the role of others at the airfield too, even where there is already a collaborative decision-making system (A-CDM). Thanks to digital towers and AI, controllers’ ability to see, in an instant, the ground status thanks to the bank of cameras, and the enhanced visuals that come with them adds significantly to their situational awareness.
Taylor says that even at the better connected sites, A-CDMs don’t always provide a view of how things are really progressing. “Because what you’re relying on is that the ground handling agents, for example, are updating the target off-block times,” he notes. “They could change at the very last minute because they’ve pulled them behind, which becomes a bit of a clunky process as you get an update at the point they update the data.” Instead, cameras and AI are feeding back information by the millisecond, rather than when it’s manually updated. This, he says, provides more granular KPIs and greater awareness, thus facilitating more informed decision making.
The fusing together of data, AI and machine learning with human handlers is likely to prove to be one of the biggest developments of ATC in aviation’s almost 120 year history. Taylor believes putting people at the top of a tower surrounded by glass, and asking them to get on with it is becoming consigned to history. “Today’s approach is to give them digital enhancements,” he says.
Project Bluebird is an effort to bring all these elements together, safely, by working with controllers to understand how they make decisions so that these behaviours can be taught to AI systems. It wants to address the ethical questions that might raise too, such as “where the responsibility lies if a human-AI system makes a mistake, how to build a system that is trusted by humans, and how to balance the need for both safety and efficiency”, the Alan Turing Institute says.
The evolution of digital capabilities and AI in ATC is one that could revolutionise the way aviation is conducted. From providing decision-critical information for on-the-ground and in-air operations, to continually monitoring for risks that controllers might miss if a unique set of circumstances arise – the potential is almost limitless. The way ATC is an area continuously in flux, developing and progressing as the new technologies are introduced, and will change further in the years to come. For now, AI is a new frontier, one that will excite and raise questions as we continue on a path towards it.
The average en-route flight delay. It’s estimated that by 2035, if there is not a significant change, delays could be as long as eight minutes per flight.
Eurocontrol and Arthur D Little
The projected compound annual growth rate of the global remote tower market between 2022–27. However, it will still be relatively small, in comparison, valued at $600m.