RTwr: Where Next?

It’s an incredibly exciting era to be part of aviation. That was the general consensus amongst friends and colleagues this year in Madrid at the CANSO World ATM Congress, writes Alex Sauriol, chief technology officer, Searidge Technologies.

Having been around in the very early days of Remote Towers over a decade ago, it is particularly exciting to see so many countries not only implementing RTS systems but making RTS the cornerstone of ATM/Airport modernization. The increasing pace of RTS adoption along with the shift towards smart/digital towers affirm for me that significant useful technology disruption in this industry can happen.

That may sound like a trivial observation but if you were part of the very small group of people that believed in Remote Towers in 2006, the affirmation is anything but trivial… and hence the excitement – this year’s WAC had talk of drones and UTM… we saw the incredible achievements of the Aireon team in making space-based ADS-B a reality and of course, the acknowledgment amongst many executives that ANSPs and large airports need to have an AI strategy.

On that AI business…

We’ve been working with various AI concepts and technologies in R&D for a few years but it really wasn’t until the last 18 months or so (or the latest generation of fast GPU cards) that I could really appreciate the hype. This was when I first saw that the early performance results of our neural network based i-video processing were starting to be better than the results from our traditional algorithmic development. Short-form: a single person in a couple of weeks developed with AI what took a half-dozen people five plus years… and his version was working better.

For the non-programmers that might be reading this, let me share with you a personal history of programming and why this AI shift is such a big deal: I technically started programming when I was 5. I did this by joining my dad on weekends, going to his office and helping him and others carry boxes of punch cards to the mainframe where they could be fed and the machine programmed. Google ‘punch card programming’ and take a look at some of those images… wow right?

Not long after, I started with BASIC, a Commodore Vic 20 and code listings from the local library. And finally, C/C++, various higher level languages and little bit of assembler mixed in takes me to today. The important part of this story is that other than physically carrying boxes, programming machines to do useful stuff for people hasn’t really changed all that much in over 30 years; specifically – a programmer needs to fully understand an application, develop a series of logical rules that govern application’s behavior and then encode those rules in a way or language that a machine will process. All that frankly is the easy part or as programmers refer to as “the first 80%”. The last 20% (taking significantly longer than the first 80% depending on complexity) is testing, QA, on-site diagnostics, debugging etc…

Our Neural Network ‘programmer’ did none of this. He produced the application simply by training the Neural Network the same way kids are taught colours and letters in Kindergarten – he pointed at an image and told the machine – “this is a plane”. No 80%… no 20%… it just worked.

This AI stuff is a big deal.

The Fonz

“Jules: Nobody’s gonna hurt anybody. We’re gonna be like three little Fonzies here. And what’s Fonzie like? Come on Yolanda what’s Fonzie like?

Yolanda: Cool?

Jules: What?

Yolanda: He’s cool.

Jules: Correctamundo. And that’s what we’re gonna be. We’re gonna be cool.” – Pulp Fiction

Having a useful conversation about AI can be challenging as it can very easily veer into science fiction, discussions about consciousness, the meaning of life and Terminator 2. And this is where we just need to be cool. We need to save the robot wars conversations and replacing an entire ANSP with a cluster of servers for another day and forum. We need to be cool and stay focused on the here and now because the AI conversation is an incredibly important one to have. AI can today impact every aspect of ATM from ab-initio training to enroute and airports, planning through execution, automation and safety nets, UTM and legacy radars and so on. Even more consequential, AI can transform the way we introduce innovation into the ATM eco-system by shifting part ‘programming’ burden from a lab to a site and from programmers to operational specialists; this alone has incredible potential to shorten development cycles, reduce costs and enhance the end user experience.

This is the here-and-now of AI for ATM. Not 10 years from now. Not next budget cycle. The AI disruption is here and it is now and this is the spirit that is driving me and everyone on the Aimee development team. We are hoping to help usher in this new era by empowering organizations to develop ATM solutions, applications and AI based digital services.

Like with our approach to remote towers, we hope that with a broad, flexible and open platform approach we can help our partners safely and efficiently translate AI concepts into digital services that can be connected to existing ATM infrastructure.

It’s an incredibly exciting era to be part of aviation – be cool.

Posted in Features Tagged with:

Comments are closed.