A Conversation With Ansible Motion Founder, Kia Cammaerts

The 2022 Autobacs SUPER GT Series begins this weekend at Okayama International Circuit, where Honda Racing will aim to build upon their existing successes. Since the launch of their front-engine layout Honda NSX-GT in 2020, they’ve won eight of the last 16 races in GT500 competition.

In November, it was announced that Honda R&D Sakura (HRD Sakura) had commissioned Ansible Motion’s latest driver-in-the-loop (DIL) simulator, the Delta S3, for use in virtual testing and developing future road and racing technologies.

Before the start of SUPER GT official pre-season testing, I had the opportunity to speak with Ansible founder and technical director, Kia Cammaerts, about the benefits of their Delta S3 DIL, as well as the growth and progression that the company has made since beginning operations in 2009. It was an informative and in-depth chat, with some very fascinating and humorous stories included!

RJO: First, as an explainer for those who may not know – because this may be some reader’s first experience – what is a driver in the loop simulator?

KC: Essentially, what we do is we create the bit that sits between the real human driver and the imagined system, or the virtual system, whatever that is. And, of course, it’s a car, but the system that you’re focusing on deriving information about – it may not just be the whole car, it may be a particular aspect of it. So what you’ve got to do is take a real human and connect that to a virtual, imagined thing. And how do you do that?

In the simplest possible way, you’d take a racing game, and you have a rendering of what you’d see. And you have a simple way of putting driver inputs in – a little gaming steering wheel, pedals, something to change to change gear with, and you’re away, you can do that. But that doesn’t actually result in anything more than exercising the imaginary system, the virtual model. It doesn’t really engage the human. And the benefits of engaging a human, particularly a skilled one, are completely lost, so you basically sort of wind the knob up and continue winding the knob up and you get to 11.

And then that’s where we come in. We put huge emphasis on vestibular stimulation of the person. Basically, we make a motion system that will move the human around, move the driver around and give them inputs that make them behave more or less as they would in the real car. And then there’s a whole host of other immersive technologies and things that you do to make the driver react more so they’re in a real car than as though they’re in a simulator.

RJO: Ansible has been in this industry of developing simulators since 2009. What has been your greatest takeaway in the way the sector has grown in the time since you’ve been involved?

KC: When we started, driver in the loop wasn’t really a thing, except in a couple of Formula One teams. And the spur to us actually coming into being was activity in the imaginary side, the virtual models of everything. And my background was in vehicle performance prediction, using software, and all sorts of things – but in particular, lap time simulation. And you hit a limit, where you couldn’t improve the correlation between the imaginary model and the real behavior of the real car. And the conclusion to why that limit was hit was that actually cars aren’t magic boxes that just make lap times.

You need a driver to be able to use them, and extract the performance that’s available from the car. But, you know, a typical scenario with computer optimization and vehicle dynamics setup, is you end up with an undriveable car, that would be really, really quick, if only you could drive it, but it’s undriveable. You’ve got to make a compromise between peak theoretical performance and what real people can actually extract. And there are all sorts of quite subtle corollaries with that – for example, in performance categories, it’s okay to make a car that’s razor sharp, and will allow you to extract a very quick lap time. But what if you want to do that lap, after lap, after lap? If the loading on the driver is such that the driver is very stressed and is struggling, then the performance of the system of driver and car will drop off.

You can’t just predict lap times through computer techniques. And it’s probably a good thing that you can’t, because stuff will get quite boring if you could. You need a skilled human to actually exercise the system you’re trying to optimise, in order to find out whether [or not] it’s actually any good. And how do you do that? You have to connect a skilled human driver, with your vehicle model. And if you just give them a gaming wheel, they’re not driving, they’re just, they’re just operating a game. So, the absolute minimum is to provide really high quality of feedback on the vehicle dynamics.

Back in 2009, that wasn’t really a thing, except for a couple of F1 teams that were doing it. People started to try to do this, in the lower categories. And they were trying to achieve two things in these sort of journeyman categories: They were kind of trying to train the driver, I think, more than develop the car. And in the senior categories, they were definitely trying to develop the car, but develop the car in a way that their drivers could actually extract the performance. And it wasn’t fully understood how to do that, and what the minimum was. On the lower categories, just learning the corners might be enough in the early stages, but it wasn’t really enough to develop the drivers. And in the senior categories, there was never enough.

It was just the start. People understood what you could do with this, and they were willing to try. But you actually need to put so much into it, that the benefits don’t come to you until you’ve got quite sophisticated, quite expensive, and quite large facilities – and quite a large commitment of resource into that. Over the first, let’s say, three to five years, we saw driver in the loop become an accepted part of people’s development strategy. But it was really the poorer cousin of things like asking, “Do we spend our money on aero? Do we spend it on powertrain development? Do we spend it on embedded code to improve our controllers for our various systems? Do we spend it on on test time?” In the senior categories, of course, they couldn’t spend all the money on test time that they wanted to because they were limited.

I think [the progress] came more or less [from the] top down, not bottom up. The senior categories started to realize that instead of a new aero package, if they invested in a suitable engineering quality driver in the loop setup – which was a very big entry fee, and a very big learning curve – that they would start to get benefits that were the equivalent of another aero package. And that started to trickle down. In the second half of that decade, it became an established tool, to the extent that in the senior categories – if your DIL broke, you might not just bother turning up at the race weekend, because you wouldn’t have the information required – in a close-run category where conditions are changing, and rules are driving you in slightly different directions each time – to be competitive.

RJO: Honda Racing Corporation in Japan recently acquired one of your Delta S3 DIL simulators for the purposes of training their drivers and engineers, and they’re keen to apply this to good use in SUPER GT and Super Formula competition. As I understand it, they’re not the only Japanese clients that have invested in your DIL systems?

KC: We have multiple customers in Japan for our top tier driver in the loop products. In general, any customer in any category can choose to either spend less on physical tests, if they have a high end DIL, or they can spend the same on physical tests, but derive far more value from it. Because using DIL appropriately allows you to either search a larger solution space than you would otherwise contemplate – or, once you’ve isolated an area of interest, go into far more detail in that area then you could, given finite test time. For a particular category, if test time is rationed, DIL allows you to do far more testing – albeit, in the virtual domain with pluses and minuses compared to the physical domain. If you’re [needing] to economize, then it could certainly allow you to economize on physical tests. But I’m not sure many senior categories would would take that choice!

RJO: I love hearing a good pitch, so have the floor – what are some of the benefits for a manufacturer or team to invest in the Delta S3 DIL simulator, in terms of driver training, and track acclimatization?

KC: Driver training is a shoo-in, essentially. With a professional engineering quality DIL, you have a lab-like environment, that reflects the detail of your vehicle. If you have a really good representation of your vehicle, and you have a really good driver in the loop system, then what you can do is exactly repeat particular conditions – so a driver can drive exactly the same car, in exactly the same state, over and over and over again until they fully understand it. And that’s something that just isn’t possible with physical tests, because the car is changing all the time, the weather conditions are changing, the tyres are changing, the track is changing. But in DIL space, you can exactly define what the test conditions are.

You can also run through a wide variety of different test conditions. And a good example would be weather – you can change the weather instantly. You can’t do that in the real world, it’s going to do what it’s going to do. You can focus on particular aspects of the vehicle. And you can, for example, simplify some aspects, and make other aspects more detailed, so the driver focuses on those. It’s a training tool. It’s near perfect. And you never have to polish the car after an accident. It’s just an instant reset, and away you go!

I think most senior category drivers can already drive really, really well – beyond any normal human limits. They aren’t going to be trained in the circuit, but they may well be trained in the very, very complex ways of operating vehicles. Particularly hybrid vehicles and in the future, electric vehicles – we’ve got that to look forward to. With hybrid vehicles, you’ve got extremely complex strategies that have to optimise multiple aspects of vehicle performance in a way that’s very, very hard to do in a one off optimisation exercise. It’s very much a matter of how the driver actually operates the vehicle.

And racecraft is another thing. Your optimisation may be great on an uninterrupted lap, but, if you have the requirement to follow slightly slower cars, overtake them, do other racing activities – you can’t just optimize in one dimension, so the number of degrees of freedom in that system that you have to optimise is beyond a simple computer. And even the most complex computer optimisation techniques, they will not give you a winning strategy straight out of the box. You have to put your driver into your car with your strategies, and exercise those as fully as you possibly can in order to understand where your freedom to operate is.

At the race weekend, the universe will give you something that’s different to what you model. But hopefully, you’ll have modeled around it, and you will know where you should be adjusting things. Your driver will be trained in how to adjust those things, and how the vehicle will respond when they do. If we take the training beyond the sort of normal kind of, ‘how do I get this car around quickly?’ and you’re taken into that level of complexity, you’re talking about a chess strategy style of complexity. And the top teams are playing chess with each other with these tools, they’re not just going out and slugging like a boxing match. Not that I wish to deprecate the art of boxing!

RJO: We spoke about how the DIL simulator field as a whole has grown since you’ve gotten into the field. How great a leap forward has your own technology taken in that span?

KC: The initial thinking behind the first motion system we did was driven by human perception and need to stimulate that human perception at a high enough level that the driver could, to an extent, suspend disbelief. But more than that, actually gain proper motion stimulation, enough to map what they were feeling in the simulator to what would happen on track. And while it’s all a massive approximation – because a track is large, and a simulator is small, and therefore it cannot, by definition, reproduce exactly, the sensations on track – we got quite close to where you where you’d want to be as a minimum entry point. But over time, people’s appetite for machine size, motion, space size, the duration of the signals, the fidelity of the signals, has increased. They’ve been particular aspects of the motion specification that’s gained a lot of scrutiny, particularly in motorsport.

In parallel to the growth in motorsport, there’s been an explosion in road car use of DIL. That’s everything from a high performance road car, which is a very closely aligned technique to to motorsport development, but is focused at a more common sort of driver. Our simulators are targeted at both road car and racecar now. In the beginning, they were targeted entirely at race car. But the technology is very applicable to a road car. However the race car is still driving a lot of the development of these simulators.

The “push” was in particular aspects of motion. Although the original systems are hardly still in use, they’re still capable of delivering pretty good results – and far better than any [other] machines at the time. Essentially, the machines have become bigger and they’ve become faster, and certain aspects are emphasized. But the driving environments have become much more sophisticated, when they were relatively simple to start with. Now we’re packing in more and more layers of immersivity and more realism, to the extent that in a proper motorsports application, the driver – once they’re in the cabin, whether it’s a monocoque, or a sports car, or what have you – isn’t seeing anything that isn’t what they would see on track.

In parallel with machine development, there’s been consistent development in software fidelity, so you can simulate more parts of the car, to a higher level of detail. You can render the visual environment that the drivers are in to a higher level of detail. And let’s not forget that pretty much the key aspect of vehicle physics is the tyre interaction with the ground. And the way that you characterise the ground surface has improved significantly as well. From relatively coarse, gridded circuit representations, to extremely fine circuit representations with pretty much every bump, every curb, and every feature mapped, and an ecosystem around that to capture that data and provide it to the teams.

It is quite interesting that there are a number of other parallel industries that have sprung up to support DIL simulation at the very highest level, to the extent that a particular bump will make a difference on a corner, and therefore you need to know that that bump is modeled properly. It really is down to that kind of level of detail!

Overall, computational powers improved, software fidelity’s improved, there has been a much wider spread of the knowledge of how to build the software models of the various physical systems on the car. And that was a real limitation five to ten years ago. There were very few people that knew how to build models that would be any good at all. And now – I wouldn’t say it’s universal, but in general – there’s far greater understanding of the need to invest in the software modeling of the systems. And then the hardware itself as has basically gotten bigger, faster and better, as you would probably expect.

At the heart of this is the human, and humans haven’t changed. But the acceptance of DIL in the mind of the of the professional racecar driver is another key aspect of it. Ten years ago, drivers would come and they’d have a couple of reactions, typically. And one of them would be like, “What is this? Oh, yeah, I got a game at home. Yeah, let’s try this.” Like, no, it’s not like that. The other one was, “I don’t want to get into it, because I’ll be sick, and it will train me in bad habits, because it won’t react like my car will. And therefore, I will learn how to control this, or jump into my real car. And I won’t be able to control that as well as I could because I will have been misinformed.” And you essentially don’t get either of those now.

No professional driver takes an engineering quality DIL for anything other than a workplace. As with all professional drivers, their sense of humour vanishes when they get into the cockpit, and they become quite demanding! And they do so in the in the simulator, because it’s no longer a game. It’s part of their professional life. And they want to take every advantage they can, out of every minute that they’ve got in that simulated cockpit. Nobody’s really worried about getting ill, and nobody’s worried about blunting their driving skills, if you’ve got a well-developed simulator.

There’s a complicated package behind that, but basically the statement is really: The driver will understand how the vehicle is going to behave. They’ve got confidence in the vehicle modeling. And they’ve got confidence in the simulator. It isn’t going to feel exactly like their car. But everything they learn from that, once they’ve done that mapping, is going to be transferable to their car. And so they’re basically on track. And they’re basically learning and steering the development of their cars. And you know that’s something that’s really marked, the buy-in from the drivers.

RJO: What’s been some of the best feedback you’ve received from drivers who’ve made DIL a part of their everyday, professional lives?

KC: Any time a driver takes [our simulation] as literal truth is a validation of everything that we’ve done. And every time a driver gets out, and harangues the vehicle dynamics guy for the behavior of the vehicle, I take that as a personal compliment.

No driver will say a simulator is exactly like the real car. More than anecdotal, it’s now just basically a routine thing – that there’s a point at which the driver, if they’re new to DIL, just stops doubting it and starts not speaking about the simulator, the motion system, the motion cueing, graphical rendering, audio pipeline, or the immersivity tools within the cabin – but starts talking about the the simulated vehicles, starts talking about the car, and speaking to their race engineers about the car.

That’s the point at which you know, you can step back into the shadows, and you’re doing your job. And that doesn’t sound like a huge thing. But it’s a huge thing to us – because it was the whole point of doing what we did. And this is something that a driver new to DIL will take a little bit of time to settle into. And once they have, you just breathe a sigh of relief, and you think, “Right, we can we can step back, and have the drivers doing engineering with their race engineer.”

We had a driver once, during an extended evaluation over a period of a few days. He was testing a GT style car at the Nürburgring. And on the second day, he had an off, and he launched his virtual vehicle into the trees. We reset the simulation, and off he went. He carried on for about two minutes – and then he put his hand up and said, “I want to get out.” And he didn’t drive for the rest of the day. His colleagues explained to us that he’d actually had a real accident at that point [of the circuit], and had been hospitalized. And after he’d had his virtual accident, he had the same feelings. So the poor fellow was basically getting flashbacks to the real thing!

But that’s a testament to the power of the simulator. For him, it was not a simulation. For him, it was a real accident. And all of his mental faculties, and all of his reactions were the same as for the real accident. So what we’ve done is – we managed to convince the whole of his psyche that he was in a car that was behaving like a car. And that accident was more or less a real one. And so of course he had a big panic, and left after that because the Nordschleife is full of trees. And if you have a if you have an off in the Nordschleife, your car is in a tree! So it’s not a good thing at all.

That was a wake up moment for me: We’re not doing simulators for these people now. For a driver, they’re in the car. And that was when I started looking at it differently, and started listening for the change in when the drivers stopped talking about the simulator, and started talking about the car.

Images courtesy of Ansible Motion and Honda