F1 circuits designed by AI: fan opinions & stats shape new tracks


Motor racing and Formula 1 in particular has always been a numbers game, but work behind the scenes is transforming the future using simulation, data analysis and fan views

2020 AlphaTauri, Franz Tost

Data is crucial to teams but F1 itself is now using it to form its future

Peter Fox/Getty Images

Next time you comment on a race thread or forum post about a dull F1 race or big piece of news, you may well be doing more than just amusing your audience; you could be playing a part in making the series more thrilling.

Formula 1 has developed artificial intelligence that combines fans’ social media and Reddit reactions with race and simulation data, to come up with more exciting circuit designs and modifications to existing tracks.

The technology is meant to encourage close and exciting racing – rather than just focusing on a high number of DRS-assisted overtakes – and is already being used. It is also helping to shape a new generation of car regulations for 2026.

“We’re getting our first results from linguistic analysis of social media, particularly on Reddit”

The pitlane at the Hanoi street circuit, scheduled to host the first Vietnamese Grand Prix this year, was shortened after F1’s software showed that this would lead to more exciting race strategies. It is also influencing the design of new tracks in Saudi Arabia.

The technology brings the promise of a an F1 opening round which lives up to the pre-season build-up, with data being used to revise Melbourne’s Albert Park circuit.

It could even lead to the development of an all-new showcase track, the design led by data, representing the ‘ultimate’ F1 circuit.

F1 2021 Car mock up

F1’s next generation car will be the first data-driven project

Clive Mason/Getty Images

“We want fast corners where the cars get stretched because they look spectacular; we want to promote overtaking opportunities,” Pat Symonds, Formula 1’s chief technical officer, told Motor Sport.

“We’ve looked at overtaking models in the past and run driver-in-the-loop simulations but we’ve taken that to a different level with using artificial intelligence.”

F1’s algorithm takes in race statistics, including pitstops and overtaking moves, as well as tyre wear data, which is combined with a range of fan reactions.

The Vietnam Grand Prix would have been the first instance of data-based design bringing about change.

As well as analysing the language and tone of online posts, F1 has wired up audiences to sensors that monitor their emotional state, – known as galvanic skin response to help assess the excitement of any race.

“We’ve been using Galvanic Skin Response for two years [recording viewers’] emotional responses while they’re watching a race and monitoring them, seeing when their emotions are high and when they’re good and then relate that to what they’re seeing on the screen,” says Symonds.

“We’re combining that with dial testing, they’re rating on a tablet every five minutes how they think the race is going so far. It’s a much more scientific thing because it does remove some of the bias which is inevitable when it comes to customer response-type stuff.

“We’ve started a project and getting our first results now from linguistic analysis of social media, particularly on Reddit. The Formula 1 community on there is a very large one and we’re doing analysis on that and looking at the sentiments of the linguistics that are being used. Once we’ve developed the mathematics, we’ll start moving that to other forums like Twitter so we understand our different kinds of fans.”

Related article

“All of this helps us understand the sport in a way that I think has never been done before. Working with the FIA [motor racing’s governing body], we’re starting to understand the cars and their performance, the drivers, the circuits and then trying to design better circuits, the tyres and trying to understand how to make the tyres better for the sport, and above all, understanding what our audience really wants.”

Symonds says that F1 has developed the tools to turn the data into usable findings with its partner, Amazon Web Services (AWS) . “It’s easy to focus on core data but it’s all the peripheral stuff and those interactions, that’s where machine learning is coming in,” he says.

“When you’ve got huge amounts of data there’s a huge amount to be said for using machine learning, artificial intelligence and neural networks.

“We’ve produced what I think is the first-ever sophisticated overtaking model and taken that to a different level using a system that I’d call artificial intelligence. We have a very good model now, so when we’re looking at new tracks or modifications to tracks, we can give a quantitive and evidenced-based decision.”

Formula 1 is working with circuit designers to move away from an individual’s artistic vision in favour of an approach led by scientific method.

This involves building models and simulations that use the data, and take into account circuit topography, average weather and track surface, to create a track conducive to entertaining racing.

Had the 2020 season played out as originally scheduled, the Vietnam Grand Prix would have been the first instance of data-based design bringing about change. The Hanoi circuit underwent several changes from the original plan, including a shortening of the pit lane after simulations found that the original layout discouraged alternative strategies in a race.

“The original layout and pit entry was very very long and convoluted and therefore, a pitstop loss was very high — nearly 30 seconds,” says Symonds. “Of course, our fans tell us they like pitstops, strategy and races with more than one stop in them.

“If the pit loss is more than 28-29 seconds, you’re not going to have two-stop races because they won’t work out. We shortened the pit lane significantly, to a point where it’s actually quicker to go through the pits than on the circuit. It doesn’t guarantee a two-stop race, but it makes it far more likely to happen.”

Formula 1 Vietnam GP circuit, Hanoi

Vietnam’s Hanoi circuit changed the pit lane length to make multi-stop races more likely

Manan Vatsyayana/AFP via Getty Images

Saudi Arabia will have one of the first new circuits to use the approach. “There’s going to be a street circuit used for a few years and then a more permanent circuit is being built,” says Symonds.

“We’re working very closely with the circuit designers, Tilke, Apex, Alex Wurz’s group for the circuit in Saudi, we’ve established a very good working relationship, particularly with Tilke where they really do understand some of the things that we want that, to be honest, just never occurred to people before.

Symonds says the team has already highlighted more aggressive track surfaces as a way of improving the spectacle, with a harsher surface interacting with the top layer of the tyres in a way that would encourage tyre wear and open up strategy options.

“It’s a bit of a shame that we bookend quite an exciting F1 championship with two rather poor tracks”

Rob Smedley, the former Ferrari and Williams engineer who is now F1’s director of data systems says that F1 isn’t looking for “manufactured overtaking” at new circuits “but something that the driver has to work really hard … options that the drivers can take to effect an overtake.”

As an example, he cites this year’s race at Mugello, which many had thought would be a procession. Smedley says that engineers saw it differently.

“There was a lot of overtaking there and the reason was because of the odd-camber corner [Turn One], with more than one line around there,” says Smedley.

Related article

F1 builds simulator to trial new race formats

F1 builds simulator to trial new race formats

Formula 1 will begin trialling new race weekend formats — such as a reverse-grid qualifying race — from the end of next year, using a purpose-built simulator. After facing opposition…

By Jake Williams-Smith

“The camber increases the more you go to the outside of the corner and you can take the line around the middle of the corner and be equally as fast as taking the inside line of the corner due to the lateral acceleration you generate.

“That’s what we’re doing in Formula 1 in conjunction with the FIA; designing new circuits or looking for new venues, circuit layouts, taking that engineering corner knowledge and trying to get that into areas of the circuit where it will promote overtaking.”

The ongoing work isn’t just restricted to new circuits. Albert Park has been a mainstay on the F1 calendar since 1996 but the circuit has seen minimal change from its inaugural year to current configuration. That is set to change in 2022.

“Melbourne is doing an upgrade for the 2022 race and we’ve been engaged with them for some time on that and looking at it,” says Symonds.

Lando Norris, Antonio Giovinazzi, 2019 Australian GP

Albert Park will be the next circuit to undergo changes with the help of F1 overtaking data

Morgan Hancock/NurPhoto via Getty Images

“It’s a bit of an unfortunate thing with the Formula 1 championship that we don’t start on the best circuit and we don’t end on the best circuit and there’s a bit of a shame that we do bookend quite an exciting series with two rather poor tracks, so we were very keen when Melbourne said they had some money and wanted to do some improvements.

“We have suggested things using our simulations and some changes which they’ve accepted.”

Theoretically, the perfect Formula 1 track to facilitate exciting racing could be designed and built, based on the numbers, according to Symonds.

“I think we could yes,” he says. “You always get an amount of interpretation in that as to what do you actually want, but yeah there’s no reason why you couldn’t. I think there would be some interpretation as to whether it had been done correctly or not but there’s always something you have to follow.”

The circuits are just the start for F1’s new modelling data, and the process is now being used to identify changes to the cars that would make racing closer. -initially with the regulations being introduced in 2022, and then with the next change in 2026.

“We’re considering the 2026 car,” says Symonds. “I won’t say we’ve quite finished on the 2022 cars, there’s still a few little bits of regulations and things that we’re trying to exploit, but we’re trying to think much more ahead to the 2026 car.

“A phrase I use with all my guys is ‘evidence-based decision making’. What I don’t want is some of the knee-jerk reactions of the old days, ‘let’s make the car’s five seconds a lap quicker because it will be better’. Everyone goes off and spends tens if not hundreds of millions of pounds building a car that is five seconds a lap quicker, and that wasn’t really the nub of the problem. Now what we want is evidence that informs our decisions.”