Good morning, Vincent. Thank you for having me. Nice to be on the podcast. Yeah, thank you for the introduction. My name is Sebastiaan de Stigter. I'm Dutch. I'm a project manager at LVNL, working on a number of projects in the systems and infrastructure and domain. One of the projects was the introduction of the time-based separation at Schiphol, which we'll discuss. I started at LVNL just before the COVID Pandemic hit. I was very lucky to make that switch just before in the beginning of 2020 I started and at that moment I returned from working abroad for over 10 years. I had a number of positions before moving to LVNL. So yeah, let's say in short, I obtained my master's degree in aerospace engineering in 2005. Started my career in a small aviation consultancy in the Hague, or at least by then it was small.
It was called to70. You had the managing director Ruud Ummels in season two on your podcast. He was employee number seven. I think I was employee number nine and I'm think in hindsight, that was a great start in my career. So it was a small company, large ambitions. I worked there for five years with various clients in various places of the world and I learned a lot that, yeah, so that solidified also my love for aviation and air traffic management. Then my wife got an offer in South Africa, so I decided to move with her to South Africa, try to get into ANSP of South Africa was difficult. Then ended up in a very small specialized company dealing with satellite tracking of assets, trains, planes, and automobiles using satellite only, not depending on terrestrial infrastructure, which is nearly not available in sub-Saharan Africa. Interesting peoples and skills.
I was there as a operations manager and doing a lot of analytics of all the data we got in just a mix of technology and skills. We had microelectronics embedded programming, hardware design, backend front end and web development. Also there I learned a lot. Then I was asked by the Max Planck Institute for Biological Cybernetics, something completely different in southern Germany to work as a new research group leader in the development of motion drive algorithms for simulators based on human perception models. Also, they're very multi multidisciplinary, working with neuroscientists, roboticists, mathematicians, psychologists, aerospace engineers, and working with a lot of real time systems. Very interesting. And then before moving to LVNL, I had a commercial and technical role at Telespazio, which was a company based in Darmstadt near Frankfurt, who originally was working mostly in the space domain, an industry, an industry company developing ground systems.
And when I came there I immediately saw that there was a lot of analogy, especially on the technical side of things, between the systems used in the space industry for controlling satellites and the air traffic control systems we use to make ATC possible. Yeah, I ended up in a business development role, did that for seven years. Quite successful. The company really branched out into aviation and ATM. But for me personally, after seven years being too far away from the subject matter, in my opinion, I really was looking for a role where I could really also get closer to the subject matter again and that's how I ended up at LVNL.
Well, knowing you I would not call it ending up. I'm sure you will bounce around and in different positions at LVNL and it's really nice to see people with a multiple careers and multiple aspects and not just focused on just one thing. So congratulations for all of that. I think it's interesting, we both mentioned Covid in the intros because the subject for today will be time-based separation approaches, and we'll come to that in a second. Was that on the menu already for you when you got hired at LVNL and did they get that through or was that one of the possible things? Because it's very interesting to see an ANSP at the beginning of Covid starting such a project and keep it through it.
Yeah, so I think the project started before I joined. I took it over and at that point, I think in 2019 it started, COVID was definitely not on the map. Then of course Covid did become an issue and it did account for some delay in the project. It was more difficult to get people together. It was also more difficult to get people available for training. So on the one hand, you're trying to get a system in place which will squeeze just the last bit of toothpaste out of the toothpaste bottle, so to say, in terms of capacity. And you're in a complete slump when it comes to volume. So I think at that point we were past the point of no return and knowing that aviation will always rebound there, I think there was no moment that there was no moment that this project was ever going to be considered to be shelved. So we carried on and we also made use of that extra time to make some final tweaks and some further tests, do some more elaborate shadow mode testing. And I think we made very well use of that time. And then after the whole Corona mess ended and very soon thereafter, we actually managed to get the system into operation in January, 2023.
Yeah, great decision by LVNL leadership, not to just crop that project. Yeah. Okay. Now let us go into TBS itself. So normally to put that in very simple words, aircraft are separated by a given distance depending on their wake turbulence category and so on. And with time-based, this is one of the things that change. Can you try to explain that change in a, let's say, not too expert way at first?
Yeah, great decision by LVNL leadership, not to just crop that project. Yeah. Okay. Now let us go into TBS itself. So normally to put that in very simple words, aircraft are separated by a given distance depending on their wake turbulence category and so on. And with time-based, this is one of the things that change. Can you try to explain that change in a, let's say, not too expert way at first?
Yeah, so with time-based separation, what you are trying to improve is your performance, especially in headwind conditions. So if you are spacing aircraft at a fixed distance from each other and you have a lot of headwinds, then the interval in which aircraft cross the threshold, it decreases. So what with TBS what you try to do is irrespective of the wind conditions, try to get an equivalent time interval, which you would have in a no wind condition with distance-based separation in place so that you mitigate this effect of headwind. And especially for Schiphol being close to the coast, you'll have almost always wind. So this is definitely improving things. Now, of course for an air traffic controller, that makes it quite difficult because a distance on a radar screen you can very easily observe and control towards, and here you have a more dynamic system which changes also based on the wind conditions. So for this concept to work to space aircraft in approach using time-based, you would need to support the air traffic controller with additional cues, and that is what the TBS system amongst others provides.
So you have two things in place here. You have one TBS system which does calculations, and we'll come to that in a minute, but you also have to change and update your radar display screen, correct?
Correct. So in the end, the air traffic controllers provided it with support by means of indicators on the radar screen telling the air traffic controller what the optimal distance is between a leader and a follower aircraft. That is determined by one of three constraints. It's the biggest, of course, is wake turbulence, which we will definitely talk about because it's probably the factor which makes it difficult for the system also to be certified. Then we have the effect of radar minima that always need to be respected. And last but not least, we have the cases where heavy aircraft need more time to vacate a runway, and also that constraint has an impact on the separation between aircraft approaching the runway. And that is indicated that the intelligent approach system, the NATS system, which allows for time-based separation all takes us into account and then presents the air traffic controller with indicators to help him or her vector the aircraft and sequence them correctly.
Okay. So we have on one side the new server you bought from NATS. That's also something we'll touch in a second. And you had to adjust the HMI. Correct. What were the high level goals that you got from LVNL management? Well, was it something like with that project you need to squeeze three, four hours, sorry, two or three, four aircraft more per hour, or was it, go for it and try to see, because the whole approach is pretty new. I mean it's in operational at Heathrow and a few others. Yeah. But did you have a real hard number to reach or was the objective to increase capacity?
No, I think there was definitely as part of the whole ex-ante analysis before getting this project running, and also the whole business case of course expected an increase in capacity, or at least in terms of throughput on the runway, especially in adverse weather conditions. So what you now see, because in the original situation, you would lose capacity with deteriorating meteorological conditions that would cause delays and delays would of course then also result in costs on the airline side and to especially improve the loss of capacity, which you had, let's say on the strong headwind conditions or worsened metrological conditions, these wanted to be improved. So to improve the sustainability of our, let's say, declared capacity. So yeah, of course based on simulations there was an estimate made on what the effect was, and that was then the basis also for the business case to green light this project to be realized.
Okay. One side question here. I know this was not planned, but it's an interesting one. Was it a project purely by LVNL or was the airport and was maybe KLM as the dominant airline in Amsterdam also involved, or was it really just purely LVNL?
No, it was purely LVNL. Of course we have a very good relationship with our stakeholders, the airport, KLM as the main carrier, but also the other airlines. They were informed early on what we were going to do and there was a very lot of effort put towards keeping them in the loop in the end for both the airport and the airlines, not much changes. In the end, our performance in providing ATC services improves, but from a procedure point of view for the pilot, nothing changes. And the airport in the end airport and airlines, which simply reap benefits, but there's no active participation needed other than their buy-in.
Okay. So now coming back to that server that you bought from NATS, and they probably had to do some adjustments, the TBS concepts is partly realized using artificial intelligence. Can you tell us a bit more about this part of the thing?
Yeah, I don't know what I should call it artificial intelligence, what it is. It is a system that is trained, trained on a lot of surveillance data to train a model to, yeah, let's say measure the deceleration profiles of aircraft in different wake turbulence categories such that when the aircraft are put on the approach fix, then the compression towards the runway is such that when you cross the runway, you still maintain the adequate amount of separation maybe as a sidestep. So when we introduced RECAT, when introduced time-based separation, we also made a switch from the ICAO awake turbulence categories to the RECAT EU categories. That allowed us for a more fine grain categorization of the aircraft and also, let's say a better distinction by the intelligent approach system from NATS to calculate these compression. So the system is trained, that's all delegated to NATS. So we don't have any oversight on that.
That is also a bit of the black box for us, but we understand that it's trained on surveillance data, then profiles are being calculated, and then to validate that model, the surveillance data is simulated using the IA system to validate that all safety criteria are met. So it's a tuning of the model. I wouldn't call it AI. I think once the system is tuned, it's completely deterministic. So otherwise it would be, I think very hard to validate if you can't, yeah, if there's always a risk of AI doing something that you do not expect. At the same time, now that the system is operations, it continues to be monitored, to see how the aircraft that are approaching Schiphol, how they be behave with respect to the calculated profiles, which are at the basis of this system on the one hand to verify that all safety criteria are met, but also maybe to understand maybe there's a room for improving efficiency to further decrease separation between leader and follower. And that is an ongoing process and that is not something that you simply observe and then implement. That would require a real change process to get that implemented. So that's not done every other week, but on a regular basis, a number of times a year,
AI or machine learning has been a topic recurrently on our podcast. I mean we had discussions a few weeks ago with people from Charlie, we had Andrew Taylor from NATS on Digital towers, and what you said is very interesting because it is a black box, not in the sense of because AI, it is a black box in that case, but because you take it over from NATS and they probably don't give you all the secrets and the details, but this is always the same thing when you have a black box you have, well first you want to look into that because you like ATM, but secondly, the regulators are typically not so happy with black boxes. At which stage did you brought your regulator on board and how did they react to the project?
So yeah, we have a good relationship with our regulator, so we have them take part in the whole process of us getting this system in Schiphol. The good thing I guess is that yeah, although it's maybe quite an innovative system is not novel. So the system I think was developed by NATS starting 2012. It was part of a also supported by SESAR there. They did a lot of work on the whole wake turbulence situation doing LIDAR measurements of I think over a hundred thousand approaching flights on the various conditions. And that already building forward on that, that already takes away probably the biggest concern when decreasing separation, either by moving to the RECAT categories or by switching to time-based separation, is the effect of the amount and the severity of the encountered wake turbulence events. So for getting the system approved by the Dutch NSA, a safety argument was based first of all on the existing RECAT EU safety case, but also then on the TBS safety cases that NATS presented to the UK CAA for the operation at London Heathrow.
So when we went to our NSA, we made a safety case argument on a so-called goal structured notion that means that we build on existing safety cases. Then in addition, take onboards the seven years of operational experience that NATS had, as well as making an analysis how the Schiphol situation compares to the London situation and then to what degree then that is comparable. And of course NATS helped us very much with that and that I think made our life quite easy and also gave the supervisory authority a lot of confidence that we were not acting like cowboys, let's to say.
Yeah. So you mentioned NATS couple of time already, and that's something I'd like to dive into a bit. Who was the organization between NATS and -LVNL? Who was responsible for which parts there?
So NATS in the end is a supplier of the intelligent approach product and system. So they are ultimately responsible to provide LVNL with a solution that met the requirements LVNL put forward and delivers on the promises that they made. They were our single point of contact, so they were responsible for the project management on our side and of course brought a lot of operational expertise on board. So they had a lot of ATCOs. Also, they helped in creating the Schiphiol CONOPS, which is of course a derivative of the CONOPS at London Heathrow. They worked with our air traffic controllers in simulator trials to get them familiarized with the system, get also help us get buy-in and supporting the development of the safety case. For instance, of course NATS is not a system company, so they rely on technology partner Leidos who in the end do the development and maintain the system and provide adaptations and custom integrations of course.
So that is a very important partner. And then the third party involved also under let's say subcontracting by NATS is Think Aviation consulting. And they worked very much with tuning this model, the so-called ORD profiles, optimal runway delivery profiles to tailor that to the Amsterdam situation. For LVNL, we were pretty much then responsible for everything else, which is a huge amount of things that we need to do to have such a change in place. So of course the technical side of things, getting the systems then also let's say physically integrated in our test acceptance production environment, but also in our training simulator. In the end, we are responsible for the safety case. So we need to build that, sure getting approval. We need to design and execute the whole training program to get the air traffic controllers trained and confident to work with the system. Also, technical trainings. We have first, second, third line technical support staff that needs to maintain the system. We need to communicate to the stakeholders airport, airline. As mentioned before, we need to prepare the whole monitoring of the system post operations. So we have ongoing calculation of KPIs, monitoring this, the system, the operational performance. Yeah, I think that in a nutshell, I probably forgot many, many aspects, is how the work was divided between the parties.
Did you feel some kind of a difference maybe in spirit or in the way of working because NATS is an ANSP versus working with a pure system provider, was it a bit easier? Was it harder in some cases? Was it just different or are they acting like an industry partner?
A bit of both, I guess. So first, I think especially for such a big change that we did, I think it is absolutely great to have NATS as a supplier of a system because they are looking at this from a very operational perspective. They have the ATCOs part of the project who talk to our ATCOS, having air traffic controllers who have used the system extensively, who are involved in the development of the system, sitting shoulder to shoulder with our air traffic controllers in a simulator discussing working with the system in practice is something that is very hard, I think, for an industry partner to provide. And that is something which helped us, helped me as a project manager. So to get this buy-in, in the end, the operational acceptance of that, this system is something that will help us with what we're trying to achieve and this will work.
So yeah, I was super happy with that and not only in convincing the controllers, but also working with our safety department for getting the whole safety case together. Yeah, that's definitely a step up from working with an industry partner. On the commercial side, I think maybe we are the third, the second client having an installation of this system outside of the uk. We were just beat by Toronto by a couple of months, and I think there maybe, yeah, NATS is not so well prepared yet. Of course, as a client, we have many, many wishes and you would see my experiences from an industry partner, there would be more eagerness and responsiveness from to get this done, whereas we felt a bit more hesitation there. But all in all, we got it working in very good spirits and we're very happy. So yeah, I'm happy that we did this together with NATS.
I appreciate you being open and honest about that. And it's probably part of being Dutch. I know Dutch people tend to be very direct in positive as in negative, and I, I'd like to use that a bit. There were some people at LVNL that posted on LinkedIn about starting of using TBS approach as being the smoothest transition they have ever experienced unquote, was it really that nice and that easy or were there some rough patches along the way?
And once we took the system operationally, things indeed went very smooth. So yeah, when we started in the morning of January 26th when the system was put into operation, we had a perfect meteorological conditions to test system on the one hand. On the other hand, it was, so it was stormy weather, there was strong westerly winds. All traffic had to be arriving on one runway rather than two, gusts up to 40 knots. And so it was really putting the system through a test. We had "prepared for the worst". We had first, second, third line, NATS, Leidos, NATS air traffic controllers in rotation plan for 48 hours to deal with anything unexpected. But yeah, it is true. In the end, we were pretty much twiddling thumbs and high fiving each other because the system was working very well. So no real rough patches whatsoever of what we expected.
And what we did see is in the beginning, a number of pilots checking in the first days through RT just to verify separation because yeah, getting TBS and then RECAT in that... Yeah, that meant that in some cases where normally separation would be five nautical miles, all of a sudden it's down to three and a half. And yeah, we had a couple of cases where they were just confirming that everything's according to plan. So that was, other than that, we had one case in the first 48 hours, which is maybe funny to say, which was not really a rough patch by any means, but we did have that the system unexpectedly during a low traffic volume switch back to distance-based mode. And maybe it's, that was a bit of a head scratcher, but within five minutes it was determined that this was correct system behavior, and maybe I can shortly dive into it, we have probably the audience here, enjoys probably also the details here.
So for time-based separation to work, of course you need to have knowledge also on the wind conditions, on the final approach path. And for that, the intelligence approach system uses mode-S downlink information of aircraft flying towards the runway to create a wind model. And in various segments, which between eight nautical miles out and the threshold wind is assessed and the wind model was populated, and during the night we expect traffic to be so low that the wind model becomes still, the information is old, the system says I can't work with that, so I will switch back to distance-based operation, normal procedure, part of the way of working with the system, however you expect that in the night and not during, let's say low traffic situations. And what we did see in this one occasion is that because there was such low traffic ATCOs were vectoring aircraft into short final, just to save a couple of track miles, and then this last segment, eight nautical miles out was not populated with mode-S wind information.
And then of course, if one segment doesn't have any up-to-date wind information, then the system, the wind model becomes still and the system moves back into distance base mode. At that moment, the supervisor on duty was one of the air traffic controllers, which was part of the project, immediately thought, ah, that's probably because of that, and then instructed the ATCO, the next aircraft to just go in long. And of course the system moved back to time-based mode. And that I think, yeah, it shows that these unexpected things can happen. These are typically not the things that you test because it's just the way people work. It's not necessarily the standard procedure, but that gave us a bit of a chuckle. But in the end it was it. So it's now solved in a briefing. So if it happens, everybody understands why and how it's resolved. But that's a bit of an anecdote I can share with you.
Well, that's the difference between simulations and real life, I guess, and totally recognize this kind of things. I realize you mentioned the 26th of January, I think. So you are almost six months in. Yeah. What is your look back at the operation for now? Do you have ideas how much capacity you won? I mean, Schiphol is not the simplest airport ever. You have I think six runways, if I'm not mistaken, different concept of operations. Do you see now the actual benefits of the system in term of capacity?
When it comes to the capacity during let's say strong headwind conditions, there is undoubtedly a big benefit that we see. So we see cases where we handle three to six aircraft an hour more than we used to. So that is great news for the general sustainability of our capacity at the airport. For some people they say three to six landings an hour. How is that great? Maybe marginal to some, but I can tell you especially under those conditions, those operational conditions where airlines suffer, let's say bigger delays, this is a huge, huge benefit in terms of cost savings for the airlines punctuality. So there we see definitely benefits. If this means that we are also going to be able to increase our nominal capacity, I think it's too early to say, it's very hard to make an A/B comparison because you don't have a day in 2022, which is exactly the same as a day in 2023.
So yeah, we now have a midterm evaluation report, which came in after six months looking at a whole range of performance areas, safety, efficiency, environment, looking at initial approach speeds, final approach speeds, number of turbulence, encounters, throughputs, use of the secondary runway, et cetera. But I expect this will take probably another half year for us to be confident to increase the capacity. But yeah, the good thing is that everything is as expected in the in line with the ex-ante analysis, maybe even in these conditions, maybe even better than expected in the strong headwind conditions. So the confidence is there and yeah, I think everybody is very happy with the system. I don't think we would want to go back, sorry,
Three to six aircraft more an hour. Sounds like a lot to me at least, even if it's a multiple airport. Now, before going in our classical 5 and 50 years question. I have one more. There is a slightly more complex version of the system that is called pairwise TBS, where you move basically from the recap EU separations to a very more specific... almost aircraft type specific version of it. I've learned some things from EUROCONTROL and others.. You have something like a 96 times 96 matrix telling you, if my leader is that type of aircraft and my follower is that other type of aircraft, you can squeeze a bit more out of it. Is it something you are looking at or are you fully happy with the RECAT EU version?
If there's a potential to further improve things, I think we're generally very open to that. So we now have the current system half a year. I think we'll have it still for a while just to further work with that. We know of course that the product roadmap from NATS for the intelligent approach system does have this pairwise system on the roadmap, and I think it's coming also fairly soon, course of next year or the year thereafter. Yeah, we will need to see it's, it's too early to say there. They claim that it does have benefits. So you mentioned one or two aircraft an hour more with everything. There's a cost benefit analysis at the basis of all the changes that we make, which is not only about capacity and the money, but also the change itself. The change itself is a "disruption" in many ways. And I think we, we have many, many other things at LVNL also, which are in the pipeline. So it's up to the management to decide when this will be adopted.
Okay. Thank you for this. And now to close up our traditional question, how do you see the evolution of ATM in 5 years, but also in 50 years, when we will both be retired, but maybe still observing it.
Let's hope we still have the opportunity to observe it. Yeah, so I'm not a visionary futurist by any means. So these are my personal views based mostly on gut feelings. So yeah, that should tell you how serious you should take it. Maybe just to start with AI. Yeah, that's a bit of the elephant in the room now. I always, yeah, I'm changing my mind on AI from day to day, I feel. So I see a lot of potential. I also see a lot of hype. Yeah, I see it. I don't see it creeping into the controller working position anytime soon. I do see that AI is going to probably in the next five years, take a huge role in decision support systems, pre-tactical decisions, wether it's runway planning, sector planning, flow control, these things. We at LVNL have made first steps at that, but I don't see it taking over the tactical control of the air traffic controller at the moment.
And to be honest, not in the future necessarily. And then unmanned traffic systems. I think that's another thing that I also seem to change my mind on from time to time. I don't subscribe to the pipe dreams of pizzas being delivered on your doorstep by drones. Yeah, I have worked too long also on the aviation noise nuisance dossiers in the past to understand that this will be a nightmare if you have lousy machines dropping pizzas off at the doorstep. What I do see is that the whole energy put towards development of the UTM system, I think is very valuable, and I think it will pave the way for a new and highly automized ATN system, which in time I think will actually simply replace the current ATM system. I don't think in 50 years we will have ATCO's vectoring aircraft like we have today. So the role of the controller will undoubtedly change in a more supervisory role of monitoring an automated system making decisions, keeping the human in the loop, allowing the human to break the loop or close the loop. And then of course, it will then allow for other concepts like single piloted aircraft or optionally piloted aircraft or even personal air vehicles. Although even there, I'm a bit skeptic, especially from the social acceptance when it comes to noise. So I think that this is sometimes maybe overlooked.
Yeah, indeed. Sebastian, thank you for...
Maybe as a last...
Please, please go ahead
And then as a last remark, I think in 50 years from now, and I probably not everybody would agree with me or would be happy with me stating this. I think there's not room for national ANSPs like we have it currently. So the airspace is very fragmented. I would expect in 50 years from now, which is a long time, air traffic will be managed on a continental scale, if not global scale.
Yeah, I agree on that. I see a lot of political aspects into that and strategy and things like that, but that's probably matter for another episode. So Sebastiaan, thank you for being our guest today, for taking the time and for sharing so many insights about your project at LVNL with TBS and also your general view on ATM.
Thank you Vincent for the opportunity to have me on your podcast and hope to talk to you soon.
So you have two things in place here. You have one TBS system which does calculations, and we'll come to that in a minute, but you also have to change and update your radar display screen, correct?
Correct. So in the end, the air traffic controllers provided it with support by means of indicators on the radar screen telling the air traffic controller what the optimal distance is between a leader and a follower aircraft. That is determined by one of three constraints. It's the biggest, of course, is wake turbulence, which we will definitely talk about because it's probably the factor which makes it difficult for the system also to be certified. Then we have the effect of radar minima that always need to be respected. And last but not least, we have the cases where heavy aircraft need more time to vacate a runway, and also that constraint has an impact on the separation between aircraft approaching the runway. And that is indicated that the intelligent approach system, the NATS system, which allows for time-based separation all takes us into account and then presents the air traffic controller with indicators to help him or her vector the aircraft and sequence them correctly.
Okay. So we have on one side the new server you bought from NATS. That's also something we'll touch in a second. And you had to adjust the HMI. Correct. What were the high level goals that you got from LVNL management? Well, was it something like with that project you need to squeeze three, four hours, sorry, two or three, four aircraft more per hour, or was it, go for it and try to see, because the whole approach is pretty new. I mean it's in operational at Heathrow and a few others. Yeah. But did you have a real hard number to reach or was the objective to increase capacity?
No, I think there was definitely as part of the whole ex-ante analysis before getting this project running, and also the whole business case of course expected an increase in capacity, or at least in terms of throughput on the runway, especially in adverse weather conditions. So what you now see, because in the original situation, you would lose capacity with deteriorating meteorological conditions that would cause delays and delays would of course then also result in costs on the airline side and to especially improve the loss of capacity, which you had, let's say on the strong headwind conditions or worsened metrological conditions, these wanted to be improved. So to improve the sustainability of our, let's say, declared capacity. So yeah, of course based on simulations there was an estimate made on what the effect was, and that was then the basis also for the business case to green light this project to be realized.
Okay. One side question here. I know this was not planned, but it's an interesting one. Was it a project purely by LVNL or was the airport and was maybe KLM as the dominant airline in Amsterdam also involved, or was it really just purely LVNL?
No, it was purely LVNL. Of course we have a very good relationship with our stakeholders, the airport, KLM as the main carrier, but also the other airlines. They were informed early on what we were going to do and there was a very lot of effort put towards keeping them in the loop in the end for both the airport and the airlines, not much changes. In the end, our performance in providing ATC services improves, but from a procedure point of view for the pilot, nothing changes. And the airport in the end airport and airlines, which simply reap benefits, but there's no active participation needed other than their buy-in.
Okay. So now coming back to that server that you bought from NATS, and they probably had to do some adjustments, the TBS concepts is partly realized using artificial intelligence. Can you tell us a bit more about this part of the thing?
Yeah, I don't know what I should call it artificial intelligence, what it is. It is a system that is trained, trained on a lot of surveillance data to train a model to, yeah, let's say measure the deceleration profiles of aircraft in different wake turbulence categories such that when the aircraft are put on the approach fix, then the compression towards the runway is such that when you cross the runway, you still maintain the adequate amount of separation maybe as a sidestep. So when we introduced RECAT, when introduced time-based separation, we also made a switch from the ICAO awake turbulence categories to the RECAT EU categories. That allowed us for a more fine grain categorization of the aircraft and also, let's say a better distinction by the intelligent approach system from NATS to calculate these compression. So the system is trained, that's all delegated to NATS. So we don't have any oversight on that.
That is also a bit of the black box for us, but we understand that it's trained on surveillance data, then profiles are being calculated, and then to validate that model, the surveillance data is simulated using the IA system to validate that all safety criteria are met. So it's a tuning of the model. I wouldn't call it AI. I think once the system is tuned, it's completely deterministic. So otherwise it would be, I think very hard to validate if you can't, yeah, if there's always a risk of AI doing something that you do not expect. At the same time, now that the system is operations, it continues to be monitored, to see how the aircraft that are approaching Schiphol, how they be behave with respect to the calculated profiles, which are at the basis of this system on the one hand to verify that all safety criteria are met, but also maybe to understand maybe there's a room for improving efficiency to further decrease separation between leader and follower. And that is an ongoing process and that is not something that you simply observe and then implement. That would require a real change process to get that implemented. So that's not done every other week, but on a regular basis, a number of times a year,
AI or machine learning has been a topic recurrently on our podcast. I mean we had discussions a few weeks ago with people from Charlie, we had Andrew Taylor from NATS on Digital towers, and what you said is very interesting because it is a black box, not in the sense of because AI, it is a black box in that case, but because you take it over from NATS and they probably don't give you all the secrets and the details, but this is always the same thing when you have a black box you have, well first you want to look into that because you like ATM, but secondly, the regulators are typically not so happy with black boxes. At which stage did you brought your regulator on board and how did they react to the project?
So yeah, we have a good relationship with our regulator, so we have them take part in the whole process of us getting this system in Schiphol. The good thing I guess is that yeah, although it's maybe quite an innovative system is not novel. So the system I think was developed by NATS starting 2012. It was part of a also supported by SESAR there. They did a lot of work on the whole wake turbulence situation doing LIDAR measurements of I think over a hundred thousand approaching flights on the various conditions. And that already building forward on that, that already takes away probably the biggest concern when decreasing separation, either by moving to the RECAT categories or by switching to time-based separation, is the effect of the amount and the severity of the encountered wake turbulence events. So for getting the system approved by the Dutch NSA, a safety argument was based first of all on the existing RECAT EU safety case, but also then on the TBS safety cases that NATS presented to the UK CAA for the operation at London Heathrow.
So when we went to our NSA, we made a safety case argument on a so-called goal structured notion that means that we build on existing safety cases. Then in addition, take onboards the seven years of operational experience that NATS had, as well as making an analysis how the Schiphol situation compares to the London situation and then to what degree then that is comparable. And of course NATS helped us very much with that and that I think made our life quite easy and also gave the supervisory authority a lot of confidence that we were not acting like cowboys, let's to say.
Yeah. So you mentioned NATS couple of time already, and that's something I'd like to dive into a bit. Who was the organization between NATS and -LVNL? Who was responsible for which parts there?
So NATS in the end is a supplier of the intelligent approach product and system. So they are ultimately responsible to provide LVNL with a solution that met the requirements LVNL put forward and delivers on the promises that they made. They were our single point of contact, so they were responsible for the project management on our side and of course brought a lot of operational expertise on board. So they had a lot of ATCOs. Also, they helped in creating the Schiphiol CONOPS, which is of course a derivative of the CONOPS at London Heathrow. They worked with our air traffic controllers in simulator trials to get them familiarized with the system, get also help us get buy-in and supporting the development of the safety case. For instance, of course NATS is not a system company, so they rely on technology partner Leidos who in the end do the development and maintain the system and provide adaptations and custom integrations of course.
So that is a very important partner. And then the third party involved also under let's say subcontracting by NATS is Think Aviation consulting. And they worked very much with tuning this model, the so-called ORD profiles, optimal runway delivery profiles to tailor that to the Amsterdam situation. For LVNL, we were pretty much then responsible for everything else, which is a huge amount of things that we need to do to have such a change in place. So of course the technical side of things, getting the systems then also let's say physically integrated in our test acceptance production environment, but also in our training simulator. In the end, we are responsible for the safety case. So we need to build that, sure getting approval. We need to design and execute the whole training program to get the air traffic controllers trained and confident to work with the system. Also, technical trainings. We have first, second, third line technical support staff that needs to maintain the system. We need to communicate to the stakeholders airport, airline. As mentioned before, we need to prepare the whole monitoring of the system post operations. So we have ongoing calculation of KPIs, monitoring this, the system, the operational performance. Yeah, I think that in a nutshell, I probably forgot many, many aspects, is how the work was divided between the parties.
Did you feel some kind of a difference maybe in spirit or in the way of working because NATS is an ANSP versus working with a pure system provider, was it a bit easier? Was it harder in some cases? Was it just different or are they acting like an industry partner?
A bit of both, I guess. So first, I think especially for such a big change that we did, I think it is absolutely great to have NATS as a supplier of a system because they are looking at this from a very operational perspective. They have the ATCOs part of the project who talk to our ATCOS, having air traffic controllers who have used the system extensively, who are involved in the development of the system, sitting shoulder to shoulder with our air traffic controllers in a simulator discussing working with the system in practice is something that is very hard, I think, for an industry partner to provide. And that is something which helped us, helped me as a project manager. So to get this buy-in, in the end, the operational acceptance of that, this system is something that will help us with what we're trying to achieve and this will work.
So yeah, I was super happy with that and not only in convincing the controllers, but also working with our safety department for getting the whole safety case together. Yeah, that's definitely a step up from working with an industry partner. On the commercial side, I think maybe we are the third, the second client having an installation of this system outside of the uk. We were just beat by Toronto by a couple of months, and I think there maybe, yeah, NATS is not so well prepared yet. Of course, as a client, we have many, many wishes and you would see my experiences from an industry partner, there would be more eagerness and responsiveness from to get this done, whereas we felt a bit more hesitation there. But all in all, we got it working in very good spirits and we're very happy. So yeah, I'm happy that we did this together with NATS.
I appreciate you being open and honest about that. And it's probably part of being Dutch. I know Dutch people tend to be very direct in positive as in negative, and I, I'd like to use that a bit. There were some people at LVNL that posted on LinkedIn about starting of using TBS approach as being the smoothest transition they have ever experienced unquote, was it really that nice and that easy or were there some rough patches along the way?
And once we took the system operationally, things indeed went very smooth. So yeah, when we started in the morning of January 26th when the system was put into operation, we had a perfect meteorological conditions to test system on the one hand. On the other hand, it was, so it was stormy weather, there was strong westerly winds. All traffic had to be arriving on one runway rather than two, gusts up to 40 knots. And so it was really putting the system through a test. We had "prepared for the worst". We had first, second, third line, NATS, Leidos, NATS air traffic controllers in rotation plan for 48 hours to deal with anything unexpected. But yeah, it is true. In the end, we were pretty much twiddling thumbs and high fiving each other because the system was working very well. So no real rough patches whatsoever of what we expected.
And what we did see is in the beginning, a number of pilots checking in the first days through RT just to verify separation because yeah, getting TBS and then RECAT in that... Yeah, that meant that in some cases where normally separation would be five nautical miles, all of a sudden it's down to three and a half. And yeah, we had a couple of cases where they were just confirming that everything's according to plan. So that was, other than that, we had one case in the first 48 hours, which is maybe funny to say, which was not really a rough patch by any means, but we did have that the system unexpectedly during a low traffic volume switch back to distance-based mode. And maybe it's, that was a bit of a head scratcher, but within five minutes it was determined that this was correct system behavior, and maybe I can shortly dive into it, we have probably the audience here, enjoys probably also the details here.
So for time-based separation to work, of course you need to have knowledge also on the wind conditions, on the final approach path. And for that, the intelligence approach system uses mode-S downlink information of aircraft flying towards the runway to create a wind model. And in various segments, which between eight nautical miles out and the threshold wind is assessed and the wind model was populated, and during the night we expect traffic to be so low that the wind model becomes still, the information is old, the system says I can't work with that, so I will switch back to distance-based operation, normal procedure, part of the way of working with the system, however you expect that in the night and not during, let's say low traffic situations. And what we did see in this one occasion is that because there was such low traffic ATCOs were vectoring aircraft into short final, just to save a couple of track miles, and then this last segment, eight nautical miles out was not populated with mode-S wind information.
And then of course, if one segment doesn't have any up-to-date wind information, then the system, the wind model becomes still and the system moves back into distance base mode. At that moment, the supervisor on duty was one of the air traffic controllers, which was part of the project, immediately thought, ah, that's probably because of that, and then instructed the ATCO, the next aircraft to just go in long. And of course the system moved back to time-based mode. And that I think, yeah, it shows that these unexpected things can happen. These are typically not the things that you test because it's just the way people work. It's not necessarily the standard procedure, but that gave us a bit of a chuckle. But in the end it was it. So it's now solved in a briefing. So if it happens, everybody understands why and how it's resolved. But that's a bit of an anecdote I can share with you.
Well, that's the difference between simulations and real life, I guess, and totally recognize this kind of things. I realize you mentioned the 26th of January, I think. So you are almost six months in. Yeah. What is your look back at the operation for now? Do you have ideas how much capacity you won? I mean, Schiphol is not the simplest airport ever. You have I think six runways, if I'm not mistaken, different concept of operations. Do you see now the actual benefits of the system in term of capacity?
When it comes to the capacity during let's say strong headwind conditions, there is undoubtedly a big benefit that we see. So we see cases where we handle three to six aircraft an hour more than we used to. So that is great news for the general sustainability of our capacity at the airport. For some people they say three to six landings an hour. How is that great? Maybe marginal to some, but I can tell you especially under those conditions, those operational conditions where airlines suffer, let's say bigger delays, this is a huge, huge benefit in terms of cost savings for the airlines punctuality. So there we see definitely benefits. If this means that we are also going to be able to increase our nominal capacity, I think it's too early to say, it's very hard to make an A/B comparison because you don't have a day in 2022, which is exactly the same as a day in 2023.
So yeah, we now have a midterm evaluation report, which came in after six months looking at a whole range of performance areas, safety, efficiency, environment, looking at initial approach speeds, final approach speeds, number of turbulence, encounters, throughputs, use of the secondary runway, et cetera. But I expect this will take probably another half year for us to be confident to increase the capacity. But yeah, the good thing is that everything is as expected in the in line with the ex-ante analysis, maybe even in these conditions, maybe even better than expected in the strong headwind conditions. So the confidence is there and yeah, I think everybody is very happy with the system. I don't think we would want to go back, sorry,
Three to six aircraft more an hour. Sounds like a lot to me at least, even if it's a multiple airport. Now, before going in our classical 5 and 50 years question. I have one more. There is a slightly more complex version of the system that is called pairwise TBS, where you move basically from the recap EU separations to a very more specific... almost aircraft type specific version of it. I've learned some things from EUROCONTROL and others.. You have something like a 96 times 96 matrix telling you, if my leader is that type of aircraft and my follower is that other type of aircraft, you can squeeze a bit more out of it. Is it something you are looking at or are you fully happy with the RECAT EU version?
If there's a potential to further improve things, I think we're generally very open to that. So we now have the current system half a year. I think we'll have it still for a while just to further work with that. We know of course that the product roadmap from NATS for the intelligent approach system does have this pairwise system on the roadmap, and I think it's coming also fairly soon, course of next year or the year thereafter. Yeah, we will need to see it's, it's too early to say there. They claim that it does have benefits. So you mentioned one or two aircraft an hour more with everything. There's a cost benefit analysis at the basis of all the changes that we make, which is not only about capacity and the money, but also the change itself. The change itself is a "disruption" in many ways. And I think we, we have many, many other things at LVNL also, which are in the pipeline. So it's up to the management to decide when this will be adopted.
Okay. Thank you for this. And now to close up our traditional question, how do you see the evolution of ATM in 5 years, but also in 50 years, when we will both be retired, but maybe still observing it.
Let's hope we still have the opportunity to observe it. Yeah, so I'm not a visionary futurist by any means. So these are my personal views based mostly on gut feelings. So yeah, that should tell you how serious you should take it. Maybe just to start with AI. Yeah, that's a bit of the elephant in the room now. I always, yeah, I'm changing my mind on AI from day to day, I feel. So I see a lot of potential. I also see a lot of hype. Yeah, I see it. I don't see it creeping into the controller working position anytime soon. I do see that AI is going to probably in the next five years, take a huge role in decision support systems, pre-tactical decisions, wether it's runway planning, sector planning, flow control, these things. We at LVNL have made first steps at that, but I don't see it taking over the tactical control of the air traffic controller at the moment.
And to be honest, not in the future necessarily. And then unmanned traffic systems. I think that's another thing that I also seem to change my mind on from time to time. I don't subscribe to the pipe dreams of pizzas being delivered on your doorstep by drones. Yeah, I have worked too long also on the aviation noise nuisance dossiers in the past to understand that this will be a nightmare if you have lousy machines dropping pizzas off at the doorstep. What I do see is that the whole energy put towards development of the UTM system, I think is very valuable, and I think it will pave the way for a new and highly automized ATN system, which in time I think will actually simply replace the current ATM system. I don't think in 50 years we will have ATCO's vectoring aircraft like we have today. So the role of the controller will undoubtedly change in a more supervisory role of monitoring an automated system making decisions, keeping the human in the loop, allowing the human to break the loop or close the loop. And then of course, it will then allow for other concepts like single piloted aircraft or optionally piloted aircraft or even personal air vehicles. Although even there, I'm a bit skeptic, especially from the social acceptance when it comes to noise. So I think that this is sometimes maybe overlooked.
Yeah, indeed. Sebastian, thank you for...
Maybe as a last...
Please, please go ahead
And then as a last remark, I think in 50 years from now, and I probably not everybody would agree with me or would be happy with me stating this. I think there's not room for national ANSPs like we have it currently. So the airspace is very fragmented. I would expect in 50 years from now, which is a long time, air traffic will be managed on a continental scale, if not global scale.
Yeah, I agree on that. I see a lot of political aspects into that and strategy and things like that, but that's probably matter for another episode. So Sebastiaan, thank you for being our guest today, for taking the time and for sharing so many insights about your project at LVNL with TBS and also your general view on ATM.
Thank you Vincent for the opportunity to have me on your podcast and hope to talk to you soon.
