The challenges of fully automating air traffic management

This week I spoke at The Journey Towards Autonomy in Civil Aerospace event organised by the Aerospace Technology Institute, addressing the challenges of fully automating ATM.  

We tend to think of autonomous things as being about self-driving cars or machines doing things without any human input, but automation is something we are now becoming used to in our everyday lives. From our phones making suggestions for us, to our TVs automatically recording shows it knows we’ve watched before.  

Automation for the aviation industry offers huge opportunities, and has the potential to open the skies to new airspace users and allow us to be more flexible and agile in the services we provide. It also poses some challenges, and there are three that I think will need to be considered above all others.  

The first is safety. Safety for NATS is what we do, every second of every day. Our role is to safely move aircraft from one place to another as efficiently as we can. The safety of thousands of flights carrying hundreds of thousands of people every single day lies with us. To ensure that safety, layers upon layers of mechanisms and procedures are embedded into what we do.  

Automation can bring with it the opportunity to further improve safety levels.  And it’s already in our operation. The big jump for the future will be from controllers making the decisions with tools to support them.  to the technology making the decision without a human to check and then ‘accept’ the solution. 

A lot of time we compare the automation of the aviation industry with that of autonomous cars, but in reality, the safety levels within the two industries are not comparable. We need even more stringent acceptance criteria.  Approximately 27,000 people are killed or seriously injured in car related accidents every year in the UK alone – that’s the same number of people it would take to fill 180 Airbus A320s. In 2019, there were approx. 257 commercial aviation fatalities anywhere in the world. The level of safety assurance that will be required to implement any automation will need to reflect that additional safety level.  

This leads onto the second challenge: complexity. Airspace is complex and the way we manage it requires skill and judgement. It takes around three years to train as an air traffic controller, after a taxing selection process. The reason the human brain is so good at problem solving in this environment is because it can process a lot of information, but importantly, it can also deal with ambiguity. A machine can manage a lot more information, but not ambiguity. How do we ensure it can deal with a new scenario it has never seen before? How does a machine ensure the answer it creates is safe and efficient? It needs to be correct, 100% of the time. 

Another complexity is our neighbours, we are working with European partners to harmonise air traffic management but if the UK had a fully autonomous ATM system, and our neighbouring ANSPs didn’t it would make the interface more than a bit tricky.  

The third challenge is the human acceptance of automation – whether that’s the travelling passengers, pilots or regulators. If the human doesn’t trust the technology, then we may never see it reach its full potential. Acceptance by the passenger is important, but as we progress along the automation journey in the ATM environment, the trust between controllers and technology is essential, and that is why they are integral to the development of these technology and systems.  

The hit of COVID-19 has really demonstrated the impact on the industry of external factors, and how we must remain adaptable and flexible. A technical solution today may be obsolete in a few years. But we know automation does and will play an increasing role in supporting our controllers in providing the safest and most efficient service to aircraft flying through our airspace.