AvTalk Episode 177: Our attempt at a summer break

AVT 177 Transcript
EPISODE 177
[INTRODUCTION]
[00:00:07] IP: Hello, and welcome to Episode 177 of AvTalk. I am Ian Petchenik, technically on
vacation, and here all by myself, as Jason has gone missing. We’ll find them by next week and
be back with an all-new episode then. We would be remiss if we didn't at least touch this week
on the huge, massive incredible news that American Airlines has placed a non-refundable
deposit for 20 of Boom’s Supersonic Overture Jets.
For comment on that, we now turn to airline weekly editor and official AvTalk Chief Short Answer
Correspondent, Ned Russell. Ned, does this small non-refundable deposit by American Airlines
move the needle at all on Boom, given the fact that they do not yet even have an engine for the
aircraft?
[00:00:55] NR: Not even close, Ian.
[00:00:57] IP: Thanks, Ned. We'll have a bit more to discuss on next week's episode. Back in
March, we spoke with Andy Taylor, the Chief Solutions Officer at NATS, the UK’s air navigation
service provider, to understand the idea of digital towers and air traffic control and how the
concept has evolved to be more than just cameras and screens. We'll revisit our conversation
with Andy on this week's recap of conversations that we've enjoyed throughout 2022.
[CONVERSATION WITH ANDY]
[00:01:31] IP: Joining us now is Andy Taylor, the Chief Solutions Officer for Digital Towers at
NATS, which is the UK’s air navigation service provider, as well as a pretty forward-looking
organization from all that I've been reading about you, Andy, and all that you've been doing.
Thank you so much for coming on the podcast and talking about it today. Welcome.
[00:01:51] AT: Thank you very much. It's great to be here. Hope that I can give you some
interesting insights to what we're doing now and for the future.
© 2022 AvTalk 1
AVT 177 Transcript
[00:02:00] IP: We always like to talk about people having interesting job titles. Yours is both
interesting, and I feel like, it sets you up for having to really deliver. I mean, you are the chief
solutions officer.
[00:02:12] AT: Yes. To be honest, I think it's less what's in a title and more what you do that
you're judged by anyway. Yeah. I mean, my role, I am an ex-air traffic controller. I've worked at
many tower and approach radar units for NATS. I've also worked all the way around the world,
in consultancy roles and supporting Euro controller. I worked for them for three years on a
detachment from NATS as well. I've got a very operational background. My role actually is in
terms of working with end-users, airports, ANSPs, and getting the perfect solution for them to
meet their operational needs.
Having that operational background is where my solutions come from. I'm supported by a very
strong technical team, both within NATS technical services, and also within Searidge
technologies, who are owned by NATS, but basically, continue to be an independent and the
leading supplier of digital towers around the world.
[00:03:14] IP: Digital towers. Let's talk about what we mean when we say digital tower, because
the initial concept of the digital tower and what you're working on now has evolved over time,
hasn't it?
[00:03:24] AT: Yeah, that's true. I think, a lot of people, as soon as you talk about digital towers,
immediately think about remote towers and small airfields in really remote locations, and being
able to provide services from centralized locations. That absolutely is part of digital towers, but it
isn't the be all and end all.
For me, a digital tower basically, is where we install cameras arrays at an airport. That provides
camera data, panoramic views, which basically takes the views out of the window, which
effectively are the analog data that air traffic controllers use every day, digitizes that, such that
we can display that either at the airport. We can share it with other stakeholders at the airport,
including the airport operations team, airport operations centers, but we can also augment it, so
we can add other system data, integrate and display that.
© 2022 AvTalk 2
AVT 177 Transcript
What we've been doing most recently with Searidge, which has also been adding artificial
intelligence. Digital towers are basically a way of providing a much more powerful air traffic
management solution for airports, which can be accessed by the airport operators and other
airport stakeholders. It's quite something beyond just remote towers.
[00:04:40] IP: When you say you're adding a layer of digitization, and then adding, I guess,
different layers of technology on that, the cameras thing, that's seems to make perfect sense, is
a good place to start. Then you mentioned artificial intelligence. I was wondering if you could
expand on what those layers are?
[00:04:56] AT: Yeah. Well, like I say, the camera bit really just takes the analog data. Actually, I
would suggest that around about 70% of the data that air traffic controllers use is analog at the
moment, looking out the window. That's a significant amount of digitalization, just by that camera
data. Artificial intelligence, where we train models to use that image, video image data. Also, to
use other data sources, be that radar. Recently, we've been doing quite a lot of work in the last
two years during the pandemic of voice recognition, which again, may not seem an unusually
new subject, but it certainly is an aviation, and particularly where that voice data is broadcast
over the radio from pilots, rather than just the air traffic control users speaking to the voice
recognition, like Siri or other voice recognition on your phone, for example.
That AI is basically taking lots of different data sources, processing it in real-time, and then
providing key outputs. Now, what we tend to do is we're focused on specific problem areas.
Now, that basically means that the AI model can be trained quite efficiently in recognizing
situations and being able to provide prompts for operators to take onboard. For us, definitely,
this is about enabling the humans in the operation to be able to operate more efficiently, more
safely, because of the support that we can give them through a trained artificial intelligence
model.
I think, a lot of people are worried about AI. Suddenly, there’s some robot controller. It's not bad
at all. It's basically, for us it's, you can program an algorithm, and that can take a considerable
length of time. In our experience, and particularly the Searidge team that I work with, who've
been working on AI for seven years or so, probably longer, they realized early on that the using
© 2022 AvTalk 3
AVT 177 Transcript
artificial intelligence, machine learning techniques, they could program algorithms more
effectively and more quickly than traditional programming methods.
When we deploy these AI models, as we have done in our Heathrow lab, and we're looking at
deployments elsewhere now, those AI models are effectively trained with real data. We train the
model, and then deploy it as a closed model. It's not the AI that continues to learn while it's in
situ. It does mean, like I say, that we can get that programming effectively and rapidly deployed.
That's our approach to AI and digital towers right now.
[00:07:39] IP: Some of the problems that I'm thinking about as an AI solution being helpful, I'm
thinking about taxiing and being able to make sure that aircraft are well spaced on areas of the
airfield that might not be as visible to tower itself, or things like, runway exits and turn offs,
where you're dealing with aircraft that are already taxiing and turning off, or something like that,
making that a little bit – the increasing the situational awareness there. Are those the kinds of
things, or is this something else entirely?
[00:08:08] AT: No, that's exactly some of the use cases that we've been focused on. I think,
you've actually touched on quite a few there. Digital towels in general, even before you apply AI
can have some benefits in terms of areas that are either difficult to see, or completely invisible
from a traditional control tower. You can apply camera arrays to areas, which enables you to
effectively look through buildings that are within the line of sight from the tower. Having done
that, you've digitized that data, you can now apply AI models to it.
We've looked at untested AI in terms of looking at the turn process. When the aircraft is on
stand and being serviced, there are key milestones within that, from the baggage being
unloaded, reloaded, catering, fueling, all those kinds of items, which will have to progress before
the aircraft's ready to go again. Those kinds of events can be monitored by the artificial
intelligence. That's through the visual camera data, and also through other system data that it's
receiving. Making calls on the basis of looking all of that data, but actually being able to sort
through which is the most appropriate to use right now, so that may well be that the visual
overtakes other, maybe manually input data from other sources.
© 2022 AvTalk 4
AVT 177 Transcript
One of the other cases that you spoke about with the runway, we've been testing that at
Heathrow. We actually ran a test prior to the pandemic, where we were looking at how we could
enhance the Heathrow operation. The Heathrow control tower, its physical control tower is 87
meters tall. That means that the tower is one of the – well, it’s the tallest in the United Kingdom.
It's one of the tallest towers in Europe. It goes into cloud, low cloud. Before the operation
requires us to go to the next stage low-visibility procedures, where you have to protect the
ground landing aids, for example.
You're still in normal conditions, but basically, the controllers have a restricted view, because
they're in cloud. This is quite common for a lot of major airports where they cover a large
surface area, or they have a tall tower or both. Heathrow is both of those things. In our normal
operation day-to-day, when we have that tower in cloud, we space the arriving aircraft further
apart, because what were deployed using is a procedural control element, where we – the
controllers in the tower at Heathrow, monitor the aircraft using ground radar.
The actual aircraft object that they are monitoring is a radar return. That's not as high-fidelity as
looking at the aircraft out of the window of the tower, because you get a blob on a radar screen.
Also, it's only a one-second update. The radar updates once every second. You're looking at a
slightly historic picture, lower fidelity than looking out the window. We increase the spacing
between the arrivals just to provide us with an additional safety buffer, so that by the time the
aircraft can be confirmed as clearing the runway, and its safety strip, so that we can allow the
next aircraft to land, it takes a little bit longer than it does by looking out the window.
If we can employ a visual observation of those exits, in conditions where the tower controllers
can't see, then that was our approach. We have deployed, in addition to the panoramic
cameras, we've deployed additional cameras that overlook the runway access, rapid exit
taxiways off the north runway. That then had an AI model trained to understand what the aircraft
and particularly what it was looking for in terms of the tail of the aircraft, the tail fin clearing, the
safety strip. What the AI model was trained to do quite rapidly was to give us two triggers, visual
triggers that can be provided to the controller on out of the window video display, or alternatively,
on their radar displays.
© 2022 AvTalk 5
AVT 177 Transcript
Basically, it provides them with a trigger to say, this event has occurred. The aircraft tail has
cleared the runway safety strip. Because it's doing that, based on visual data, and the cameras
are high-resolution, ultra-high 4K cameras, so they have massive amounts of pixels. The aircraft
is looked at in terms of being hundreds of thousands of pixels. It's looking for the very last pixel
of that tail fin, clearing a line which has effectively a pixel fix. This is even more accurate than a
human controller can do, looking out the window normally. It triggers an event, say, that aircraft
has now crossed this line, assuring safety, that the next aircraft could land, roll down the runway,
and there is no chance of a wingtip to tail collision between those two, which is one of the key
tasks that we’re supplying.
That basically means that if we could prove that the system was capable of doing that routinely,
and to a high-level of accuracy, which is what we analyzed more than 40,000 arriving flights at
Heathrow doing, in all sorts of different conditions and day and night, daylight darkness,
different weather conditions, then we would be able to apply a normal final approach facing, so
not increased facing, therefore, reduced landing rate, but normal landing rates. That would
basically mean, even in towering cloud situations, we could continue to operate the normal
traffic levels, reducing delays that are impacted during that time. To basically increase safety as
well, because as I say, the level of accuracy of it when you're looking at the last pixel on the tail
of an aircraft crossing a line that's only a pixel thick. Likewise, that is incredibly, highly accurate
data that it's using. That's our concept of operation. That's what we've tested and validated so
far.
The results of it was so positive, that we actually decided that the next stage would be to look at
low-visibility operations, low-visibility procedures, in fact, where we protect slightly further away
from the runway centerline, because what you're looking to do is to avoid the tail fin interfering
with the ground landing, the instrument landing system that the next arriving aircraft is following.
Even though you apply slightly larger spacings for that, what we'd be looking at would be to try
and tease those down just a little bit, in order to on the back of that high level of accuracy of the
aircraft clearing, that line and being able to do it when you're using camera data that's running at
25 to 30 frames per second. Up to 30 times faster than radar updates, but to pixel level of
accuracy, not radar blip accuracy.
© 2022 AvTalk 6
AVT 177 Transcript
All of these things I've described previously, as any high-performing team, and I see the
Heathrow air traffic control team as a high-performing team, those controllers, just like the high-
performing teams in whatever your sport may be, those teams never just take their day-to-day
performance as being that's as good as it gets. They're always looking to try and get that little bit
better, that little bit improvement to be able to do that perfect serve, that perfect goal, whatever it
might be more often. Obviously, in an air traffic control environment more safely.
[00:15:26] IP: What's the future of the digital tower? Is there a move to expand where the AI is
helping controllers? Is this something where you need fewer controllers per shift? Or is this
something where you continue to find ways to incrementally increase the efficiency of the
towers, and that's really where you're headed?
[00:15:50] AT: There's a number of potential goals you could aim for. I think, my focus as I say
is, it's not so much from the aim to remove players from the field, but basically, to get that
performance more regularly assured and to as high level as possible. Also, to be able to –
looking at the effects just recently of the global pandemic, what we've certainly discovered is
from that is the ability to dial down, dial up operations can actually be quite difficult, especially
when you're set for high levels of performance and traffic delivery, then it's suddenly removed,
and then it suddenly starts to come back again. You want to be able to operate just as well as
you did, despite the fact that you've not been able to practice real life operations for some time.
All of that can actually be, I think, the use cases for future where you can dial it down, dial it
straight back up again. Still operate, then just as effectively as you did a few months before,
having had impacts. I'm hoping at least, we don't see these impacts again. These are use cases
that we'd never seen before. Actually, I think it is something quite relevant. Beyond that, I think
the sky is the limit in terms of what you can do with this. The key for me is the digital tower.
That's 70% of that analog data being digitized. Being able to be integrate it with other data
systems on the airport, rather than them all being standalone, where a human operator, air
traffic controller, or airport ops person, apron controller, where they have to look at different
systems and effectively processes this themselves. That integration just really helps in terms of
reducing their workload and improving efficiency.
© 2022 AvTalk 7
AVT 177 Transcript
Also, like I say, there are opportunities then to improve that, because a human can only focus
on one area at a time. We all like to think we can multitask, but actually, what we do is our brain
flips from one thing to the next. We train controllers for an extensive period of time before they
work on their own, in order to get their scan from one place to the next to the next. With artificial
intelligence and 360 degrees of coverage, visually, visual camera data, and integrated other
data sources. That AI model, or AI models can basically be looking in all different directions
simultaneously.
The controller can then be drawn to the point that the decision needs to be made. Because
certainly, humans, and I go back to those high-performing teams that we have in air traffic
control, Heathrow and other airports and ANSPs around the world. Having the controller there to
be able to make the right decision, but supported by the tool that's pointing them as to the
whether the next area is that they need to put their attention to and provide the decision, that I
think is a real game changer. Goes from reliant upon a scan, waiting for the controller to scan
back to that, now they'll do this quickly, but it's still not as quick as being done 25, 30 frames per
second, every minute of the day, regardless of whether that person has been in the seat for five
minutes, or 55 minutes, it still is working to the same level and drawing their attention to what
they need to do.
I think it's a real game changer. For me, digital towers are like the smartphone when it came on
the market. Suddenly opened up possibilities. Those applications that we started off with, we've
got hundreds of applications now that run on that smartphone. I think, digital towers are
effectively that smartphone, integrating all of these data sources, and basically providing
applications that are specific to the airport. I don't have the same apps as you've got on your
phone, I'm sure. We might share some. Basically, they'll be specific to our type of whatever
we're interested in.
I see international hubs having probably very similar applications, smaller regional airports,
perhaps maybe having some of those, or maybe having specific ones to them. I think we're on
the cusp of something really quite new and innovative for aviation and particularly for air traffic
control airport.
[END OF INTERVIEW]
© 2022 AvTalk 8
AVT 177 Transcript
[00:20:15] IP: That was Andy Taylor, Chief Solutions Officer at NATS, talking about the concept
of digital air traffic control towers and where the concept is going in the years to come. We hope
you enjoyed this little recap, and we'll be back next week with an all-new episode. Thanks as
always for listening. Until then, happy tracking.
[END]
© 2022 AvTalk 9