How Formula 1 evolved into a data-first sport on wheels

As race teams accumulate more data, Formula 1 looks more and more to predictive analytics

Where once race cars were finely tuned machines on wheels driven by fearless mavericks, today’s machines are loaded with sensors and generate massive amounts of information.

During race weekend in Barcelona, current Formula One champions Mercedes opened their garage and shared their story about how F1 cars have changed and become increasingly data-dependent in recent years.

A rise in the number of sensors and compute capabilities has seen F1 become increasingly analytical in its drive to improve racing and led to data scientists being as core to a team as mechanical engineers.

 

Driving cars is driving constant improvement

The teams’ cars are loaded with round 400 sensors each during testing season, reduced down to around 100 during a race to save weight. These sensors measure everything from basic race telemetry -- speed, location on the track etc. -- to sensors which measure tyre temperature and pressure, aerodynamic pressure on the car and suspension, as well as to internal sensors that measure and transmit information about the status of fuel, lubricant, oil, cooling systems and more. A single car can create around 500GB of data over one weekend’s racing.

“We are very much a data-driven organization,” says Geoff Willis, Director of Digital Engineering Transformation, Mercedes-AMG Petronas Motorsport. “But it wasn't an overnight switch.”

Sensors in the cars became common practice in the 1990s, and as more compute resources became available and practical for teams to utilize, they started to build simulation models to try and predict race outcomes, then build dynamic models of the car, and eventually start looking at computational fluid dynamics (CFD) for better aerodynamics. By the mid-2000s F1 cars had become the high-speed sensors on wheels that we see today, and as we came to the middle of this decade, all this information was becoming a real-time feed. 

The deluge of data is having a noticeable effect on the track. During the qualifying sessions during this year’s Barcelona Grand Prix, Lewis Hamilton’s pole-winning lap time was 4.7 seconds faster than last year, which was 2.8 seconds faster than the year before, which was another 2.6 seconds faster than the year before that. The improvement is staggering in a sport where gaps between cars in races is often measured in hundredths -- if not thousandths -- of a second. Even between races there is constant drive to improve performance.

“During the season we've probably seen 2 seconds of lap time improvement [this season]. You can't keep still.”

The improvement of lap-times comes down to a variety of factors; ever-increasing computational power means more simulation capabilities, which means car designs are more rigorously tested than ever before, ensuring the most efficient aerodynamic shape for the cars.  

The influx of data also extends to the individual races, drivers, and laps. For example, in qualifying, the Mercedes team will overlay Lewis Hamilton’s laps with those of nearby rivals and advise him where he may be losing time; braking too early or late, or taking a turn in such a way that might be inefficient. This data is downloaded, processed, and actioned in incredibly short windows; a qualifying session on the Saturday can run for as little as 12 minutes. After qualifying, the team will run millions of race simulations overnight to try and gauge the optimal strategy; predicting potential tyre wear, weather variables, when to pit to ensure clear track, and best positions etc.

Of course, the races themselves are still very hard to predict; a multi-car crash on the opening lap of the Barcelona race leading to a safety car was almost impossible to account for in simulations, adding an element of uncertainty for the teams and throwing many of their prediction models off. Although Hamilton won comfortably from pole position, his team-mate Valtteri Bottas lost a place due to a long pit stop (despite those video analytics), while a poorly timed pit stop during a virtual safety car caused Ferrari’s Sebastian Vettel to drop from second to fourth. Vettal’s team-mate Kimi Raikkonen retired from second place with an engine failure on lap 24, as did a number of other racers throughout the race, showing the need for even more predictive data around the engine.

 

More data means better racing

The Mercedes team backs ups around 45TB of data a week; this includes not only data from the cars, but videos, photos, models, simulation data, and more.

“It's grown and grown. Now we've got hundreds of channels, and as we've found the benefits of understanding what the vehicle is doing, there's been a self-reinforcing growth of data; as you understand more you want to understand more.”

The capabilities to create, store, and process massive amounts of data means every aspect of racing has changed. Away from the track, the way teams design and test their cars is changing to become a digitally-led process. And everything from impact and structural testing to aerodynamic modelling has seen teams become less and less dependent on physical testing.

“We're highly dependent on simulation for almost everything we do, physical testing is very much now a confirmation of your model prediction,” says Willis. “Going back 20 years there was a lot more experimental modelling, testing, conclusion, and evolution. Now digital techniques are giving us a lot more understanding so we're projecting further and into more detail.”

Even the pit stops -- where teams replace tyres -- are subject to massive amounts of data analysis. Pit stop rehearsals are processed via video analytics to analyze everything from how the team put the guns on the wheel nuts, to how to ensure that all the mechanisms fit together in the smoothest and most reliable way. The result is that the team is now able to perform sub-1.8 second pit stops.

“An enormous amount of work has gone into that, and even that one little -- but high profile -- task in the weekend has all of these technologies as key parts of it.”

 

Race cars are data-creating machines, hungry for more data

While there are around 60 engineers on-site during a race weekend, there will be many more staff working back at the company’s two factories in the UK helping provide additional support around the data being streamed back and forth. Even the race-drivers need a good handle on how to use information.

“Drivers are very hungry for any opportunity to improve,” says Willis. “You can't get high up in motor-racing now without a good understanding of how to use data.”

Currently there’s only minimal data measured about the drivers themselves. There are accelerometers in the earpieces to measure G-forces on the head, and there are biometric sensors in the glove that pick up the heart rate and blood oxygenation levels to comply with safety regulations.  But Willis says the team has experimented with eye-tracking during driving simulation sessions to understand not only where drivers are (and should) be looking while taking a corner, but also deciding placement of information within the car.

Few in F1 are worried about automation removing people from the sport, but utilizing AI and machine learning is high on the agenda to remove some of the uncertainty in a race. As well as constantly looking to improve race prediction models, the Mercedes team is looking to move into predictive analytics to better foresee things such as gearbox failures and accurately estimate the likelihood of overtaking other drivers in a race.

“We still rely an awful lot on coming to the circuit with our optimized set up,” says Willis, “seeing what happens, and then comparing those with the models and saying; 'we know which parameters to tune for our time model today, in this condition, on this tarmac, with this temperature, with this humidity level’.

“We would like to be able to improve that, and I strongly suspect that a lot of the clues are already there in the data we've got. We’re exploring curating that data better and connecting things that we don't realize are connected, for example in terms of environmental changes and car performance.”

 

F1 teams expanding themselves beyond just racing cars

‘Digital Transformation’ often means offering new services off the back of your own expertise, and F1 is certainly no exception. A number of teams including, Mercedes and Sauber, offer wind tunnel testing and production services, while the likes of Red Bull’s Advanced Technologies and Williams’ Advanced Engineering units offer services beyond automotive and motorsport. Williams, for example, partnered with UK start-up Aerofoil Energy to develop a device that can reduce the energy consumed by refrigerators in supermarkets.

Most ambitious of all, however, is probably McLaren’s Applied Technologies group, which is looking to take its data nous and apply it to a broad range of industries. Already a supplier of telemetry software to other F1 teams, McLaren is exploring how it can apply its expertise in use cases around healthcare and public transport such as rail. The company was at one point rumored to be an acquisition target by Apple for the potential value it could bring the Cupertino company’s automobile ambitions.

“We're building out this connected world -- the simulation and modelling -- into a variety of different businesses,” says Jonathan Neale, Chief Operating Officer, McLaren Technology Group said at the Dell Technologies World conference this year. “We're building out in digital therapeutics, we're looking at connected transport. Connected rail is an important and growing market for us. Autonomous vehicles as well; not in the AI space but in the layer below where you've got safety critical real-time software sensors and actuators which are feeding the decision-making system.

“Between 2025 and 2030, there's a good chance we'll be a business with technology as our nuclear reactor, with a series of verticals that will come and go as the markets shift.”

 

Note: This article is based on information sourced on a press trip to the Barcelona Grand Prix, hosted and funded by Pure Storage.

 

Also read:

How IBM gave the Red Bull F1 team new wings

Inside Lotus’s F1 factory

Data drives Grand Prix success at Lotus

 

Data-driven: How Mercedes-AMG Petronas F1 uses tech

Inside Mercedes' F1 factory: Simulator, data and sensors

 

Roborace: When software engineers are the heroes

Roborace's DevBot is the autonomous racecar only a few lucky developers will drive

 

1000mph land speed record quest is driven by data