1. ENGLISH
  2. News
  3. Stories
  4. 2019
  5. 10
  6. Driving at the limit with Artificial Intelligence

We use cookies (our own and those of third parties) to make our websites easier for you to use and to display advertisements in accordance with your browser settings. By continuing to use our websites, you consent to the use of cookies. Please see our Cookie Policy for more information on cookies and information on how you can change your browser's cookie settings: Cookie Policy Accept

Driving at the limit with Artificial Intelligence

Volkswagen Group Innovation engineers spent two weeks working on future-oriented autonomous driving projects at the race track in Portimao. In on the action: Norbert – an automobile prototype.

Like Walter, Norbert’s and Dieter's luggage compartments are packed full of technology
Audi RS7 – fuel consumption in l/100 km (NEDC): urban 16.1 / extra-urban 8.7 / combined 11.6–11.4; CO₂ emissions combined, g/km 265–261; efficiency class F. Photo: VW

“Norbert wants to go on the track,” blares a voice on the radio. The answer is quick to follow: “Right you are, Walter's coming in.” Shortly afterwards, a Audi RS7 with black-and-white tape drives into the pit lane and a red VW Golf GTI accelerates away onto the Portimao circuit. Norbert’s name is not really Norbert, of course: it’s Jonas Kaste.

Or to put it more precisely: Jonas is the man behind Norbert’s steering wheel. Like his colleagues Susi, Walter and Dieter, Norbert is an automobile prototype with a boot-load of computers. The two VW Golf GTI, the Passat Estate and the Audi RS7 were given their nicknames by the Volkswagen Group Innovation development team, of which Jonas Kaste is a member.

Golf GTI “Performance” – fuel consumption in l/100 km (NEDC): urban 7.5-7.3 / extra-urban 5.6-5.5 / combined 6.3-6.2; CO₂ emissions combined: 143-140; efficiency class: C. Audi RS7 – fuel consumption in l/100 km (NEDC): urban 16.1 / extra-urban 8.7 / combined 11.6–11.4; CO₂ emissions combined, g/km 265–261; efficiency class F. Photo: VW

The 25 or so engineers are meeting under the hot Portuguese sun to do some tinkering and testing: the fifth Vehicle Dynamics Convention which takes place near Faro is underway. During this two-week test phase, they are carrying out fundamental research – far removed from their offices in Braunschweig and Wolfsburg – to optimise their projects and present some of these to representatives of the individual Group brands. The latter will be able to draw on precisely this development work in the years to come. What is being tested here will generally not be available in series production cars for another ten or fifteen years.

The computer as an expert

Right, that's Norbert. A red VW GTI
Golf GTI “Performance” – fuel consumption in l/100 km (NEDC): urban 7.5-7.3 / extra-urban 5.6-5.5 / combined 6.3-6.2; CO₂ emissions combined: 143-140; efficiency class: C. Photo: VW

However, you’d be much mistaken if you were to associate vehicle dynamics with good old suspension engineers screwing on struts, dampers and multi-link axles. Nobody has hands covered in oil here – there’s not even a wrench in sight. Instead, there are lots of cables and a good Wi-Fi signal. The technicians have pushed a few tables together in the pit garage: the tools of their trade are bits and bytes. More than anything else, Jonas Kaste and his colleagues are IT specialists who use highly complicated algorithms and neural networks to try to equip the cars of the future with more and more intelligence. What does this have to do with vehicle dynamics? A great deal. After all, if the car is going to be driven without a driver, the computer has to be an expert in the field of driving dynamics.

Dieter, Walter and Norbert are able to do a lap of the Portimao circuit without the help of a security driver at the wheel. Unlike the first demonstration runs a few years ago when the Audi RS7 drove around the Grand Prix circuit at Hockenheim in Niki Lauda style as if by magic, no fixed route is saved in the system here. The cars only know the track layout: they find their ideal line themselves. In technical jargon this is known as the trajectory, and of course it depends on numerous parameters.

Nerd cap rather than overalls

This is exactly what Jonas Kaste is investigating right now. The 32-year-old mechanical engineer with the scruffy nerd cap is trying to teach the car to notice when the tyres are worn out. The principle behind this is simple: the computer calculates a route and the Golf follows this defined line at high speed. If the tyres have enough profile and grip, it hardly deviates from the specified route. But the worse the tyres get, the sooner it starts to slip and the more drift there is on corners. How does this information help? If the car knows that the tyres are in poor condition, it adjusts its strategy, allowing for a larger bend radius and reducing speed, explains Jonas. He evaluates the most important data in the vehicle on a laptop, making small adjustments to the source code.

“That's what makes the work here so fascinating: you do one or two laps and then you can correct mistakes or try out new parameters just a few metres away from the track,” says the developer enthusiastically: after a few fast laps, he clearly has as much adrenaline in his blood as a racing driver. Jokingly claiming to be a poor driver himself, Jonas only had to press a few buttons before the test drive and hold down a dead man's switch during the fast lap to make sure the prototype didn’t accidentally go out on the track unsupervised.

Engaging with colleagues and scientists

In addition to the immediate proximity of theory and practice, Jonas also attaches importance to being able to engage with colleagues working on other projects here in Portimao. By way of additional support, the Volkswagen engineers have brought in scientists from Stanford University and TU Darmstadt to help them in Portugal. “If you're stuck with a problem, you can easily get input from your colleagues here,” says Jonas enthusiastically. A chat over a grilled sausage – which the garage foreman puts on the barbecue at lunchtime – is often much more helpful than a lengthy meeting in the office.

Although the vision of the autonomous car hovers over the entire convention, much of what the engineers are investigating here with their self-driving prototypes will also benefit drivers of conventional cars. One of the engineers compares it to the Moon landing: just because it became possible to travel to the Moon 50 years ago, it doesn’t mean we go there every week. However, we still benefit from the Moon landing to this day – not least because Nasa technology gave us the Teflon pan, for example. The automotive industry’s equivalent of a non-stick coating is safety. A whole range of developments that emerge on the way to creating the self-driving car are also capable of increasing safety in conventional vehicles, helping to approach the ultimate vision of zero road fatalities. If the car notices that the tyres have worn down, for example, it can not only adapt its autonomous driving strategy but also warn the driver and request a visit to the garage.

Steering with the Playstation controller

A Playstation controller is all it takes to control Norbert and Co
Golf GTI “Performance” – fuel consumption in l/100 km (NEDC): urban 7.5-7.3 / extra-urban 5.6-5.5 / combined 6.3-6.2; CO₂ emissions combined: 143-140; efficiency class: C. Photo: VW

Developments in the area of steer-by-wire are not limited to the self-driving car either. In a VW Tiguan – surprisingly lacking a nickname of its own – Christopher Kreis of Volkswagen Group Components and his colleagues have done away with the mechanical connection between the steering wheel and the wheels, instead transmitting the steering command digitally to actuators that position the front wheels. This technology is already common practice in aircraft, but there simply hasn’t been any good reason to put the effort into developing something similar for cars to date.

But the technology is now of interest in autonomous cars – where the steering wheel is either non-existent or can be folded in and out. Here again there are definitely potential uses for cars with a driver: for example, the Tiguan can be steered not only by steering wheel but also by smartphone or even using a Playstation controller. The technology could be used in vehicles for the disabled, for example, with even more individually adapted control options that ensure even greater safety. In the future, it might also be possible to change drivers during a journey when you are tired or want to make a phone call simply by passing on the joystick.

Take over, please!

The author talks to Gamze Kabil, one of the few women at the Vehicle Dynamics Convention
Golf GTI “Performance” – fuel consumption in l/100 km (NEDC): urban 7.5 - 7.3 / extra-urban 5.6 - 5.5 / combined 6.3 - 6.2; CO₂ emissions combined: 143 - 140; efficiency class: C. Photo: VW

One of the few women in the development team here is also carrying out research in the field of steering. Gamze Kabil optimises transfer strategies, i.e. switching between autonomous and driver-bound driving. The key point here is that a steer-by-wire steering wheel does not necessarily turn when the car is driving on its own. However, if the driver wants to or is forced to intervene again on a bend, the steering angle of the wheels and that of the steering wheel no longer match. Gamze is running numerous test laps to find out how best to synchronise this, and she’s also keen to get feedback from her colleagues.

Meanwhile, steering feel without a mechanical connection is Maximilian Templer speciality: he is a doctoral student who works with neural networks – highly complex calculations that attempt to determine the optimum force with which the force feedback actuator simulates steering resistance based on learned behaviour drawn from various parameters such as steering angle, speed and a range of others factors. After all, a steer-by-wire Tiguan ought to feel like a Volkswagen, too.

What will the car of tomorrow feel like?

And while we’re on the subject of feeling: cars differ greatly in their driving dynamics nowadays. A Porsche feels different from a Skoda, driving a Bentley is different from driving a Seat. But what if the car is driving itself? If everyone ultimately calculates the same optimum line, the differences are not likely to be very great. This is exactly what the Brand DNA team wants to prevent. They are looking at how to fine-tune autonomous vehicles so as to teach them different characteristics. What will the Porsche of the future feel like? That’s not yet relevant here in Portimao. “We’re simply enabling different driving profiles to be programmed. The individual brand engineers then have to make the relevant adjustments during series development,” explains Paul Hochrein. By the time those are ready for the market, Jonas and Co. will be busy working out entirely new solutions here in Portimao.

Driving at the limit with Artificial Intelligence

Michael Gebhardt is a free-lance motor journalist. His articles are published for example by AutoBILD, Spotpress, and n-tv.

Fuel consumption

  • Golf GTI “Performance” – fuel consumption in l/100 km (NEDC): urban 7.5-7.3 / extra-urban 5.6-5.5 / combined 6.3-6.2; CO2 emissions combined: 143-140; efficiency class: C.
  • Audi RS7 – fuel consumption in l/100 km (NEDC): urban 16.1 / extra-urban 8.7 / combined 11.6–11.4; CO2 emissions combined, g/km 265–261; efficiency class F.