What is the new normal? It’s not only about the way we travel, it’s also about technology. One silver lining of the pandemic is that it drove digital transformation in many organisations at a faster pace. They simply had no choice if they wanted to continue to do busines remotely. During day four the panel discussions I attended were very focused on technology and the implications to our industry.
Charles Clancy, Chief Futurist for Mitre Labs spoke about 5G as a kick-off to the new normal panel discussion. First providing a great layman’s description of what 5G is. 5G is a different animal than 2G/3G/4G and its important for those outside of the telecom industry to understand why. 5G is all about virtualisation and edge computing – at the simplest level. So what does this mean. Today we can use software to create software-defined networks, telecommunications infrastructure and even drones. 5G allows for virtualisation of networks to support different functions. There is a term you might hear called network slicing. This means that you can literally create slices in the network to support different types of service that is required for different functions.
There’s been a lot of discussion about the use of 4G/5G for drones and BVLOS. 4G networks are built to support users who are on the ground. 5G uses something called beamforming which allows the signal to be directed in certain paths. Or potentially upward to support flying objects. 5G moves computing to the edge of the network, which means that the security of the data needs to be closely considered. When thinking about use in ATC, this is a big topic. The security of this data must be trusted. Artificial intelligence can be used to help identify potential vulnerabilities.
Speaking of Security. The next presentation I attended looked at Mechanisms for Air Traffic Security and their Integration into the ATC Infrastructure. The main direction of this discussion was looking at how threat actors could spoof GPS and ADS-B infrastructure. The result would be a plane going in the wrong direction thinking it was off course and trying to get back on course or communicating to the ATC that it was somewhere different than it actually was. I won’t attempt the detailed technical explanations, but I will say the implications of this and the apparent ease that this could be done in testing makes you really think about Cyber Safety.
Last up was a discussion as to whether or not we are ready for AI and machine learning in aviation. The short answer was ‘no’. Peter Dumont from ATCA spoke about this topic. Mainly looking at options in aircraft and the pros and cons. He illustrated an example of a simulation of a plane in severe turbulence – the auto pilot couldn’t save it and neither could the pilots in the simulator, but the AI got it back on course. Sounds like a good outcome, so why is the answer above no?
The reason is that AI learns and as it does it could change the software code and this could lead to changes that don’t align with functionality that always needs to work the same way. As the AI learns it could also take information from sources that it shouldn’t and make decisions, another concern. There is a way to ‘bound’ the AI so it cannot do this, but then he argued this is not actually AI.
Today AI is used for preventative maintenance in aircraft in order to predict possible failures of parts based on wear. It could be envisioned as an adjunct to pilots to help increase efficiency and safety while reducing the workload of humans. AI is supporting ATC regarding weather and the most efficient route topics, always with a human involved. The summarisation was that AI does not currently have a larger role in ATC.
I’d like to take a less harsh view of this. During 2020, there have been many articles in our magazine and on our website related to the topic of AI in ATC. It seems the industry might be moving a little faster down this path then offered in this session. There are also different levels of AI. Consider three tiers – one where the AI gathers information and analyzes it to give the human at the center more data than a human could likely gather and analyze on their own in a timely fashion. The second tier would be where the AI goes one step further and makes a recommendation what course of action to take. The human is still at the center and makes the final decision. It seems that these two levels already are in use and could be expanded further in the world of ATC. Level three is where the AI owns a topic from end-to-end. I agree we likely won’t see AI conducting the full job of controllers, at least likely not in our lifetimes. But this doesn’t mean that aviation is not ready for AI and that it couldn’t play a larger role.