Historically and presently, McLaren Racing is a multi-platform racing brand, currently competing in F1, IndyCar, and the Extreme E electric off-road racing series. In 2023, the team will also enter the Formula E World Championship.
McLaren is one of the most successful Formula 1 teams of all time. As of June 2022, McLaren drivers have won 183 races and 12 Drivers’ Championships, while the team has taken home eight Constructors’ Championships.
With so many years competing in motorsport, McLaren has learned plenty. And as the sport continues to grow and evolve, the team has shown an appreciation for data and developing comprehensive AI and machine learning systems to put that data to work for its cars and racers.
What’s at Stake in F1 Racing
There are more than 20 events in an F1 season, with races occuring thick and fast. The McLaren team performs testing in February and March as the race year begins, and then builds and adapts through the end of the season in November. Testing goes through design simulations, wind tunnels, and multiple iterations before going out on the track, which requires constant teamwork and communication.
McLaren aims to build AI systems that act as an extra engineer for the team to aid analysis and quickly look through data. The AI algorithms extract insights that can be used to make the car go faster.
How Machine Learning Shapes Racing Strategies
Once you’ve seen the parts in action on the track, you need to analyze them. Did they do what you were hoping they would do and give you a performance advantage? If they didn’t, why not? That knowledge is underpinned by the data and measurements the McLaren team has, and it results in building faster cars, developing them more quickly, creating better strategies, and understanding where a car is in relation to competitors — all of which play a significant role on the track.
Of course, there’s a lot of data to sift through. Most tracks have a long history, and McLaren itself is the second-oldest active team in F1 racing. Throw in additional elements such as weather, other participants with their own unique driving styles, and on-track incidents, and there are several variables that can’t be controlled. Tracking uncertainty becomes just as important as the data that’s less volatile.
There can be a lot of noise on the racetrack — and we’re not talking about the hum of the car engines! Working through noisy or extraneous data to find results that matter is a critical skill. McLaren also gets data from non-racetrack sources, such as the wind tunnel, computational fluid dynamics, test rigs, bench tests, and more. By piecing all those different datasets together, the team can quickly build upon a model and offer certain levels of confidence in specific scenarios. That data must be properly collected and labeled, however. Otherwise, it’s very difficult to generate meaningful insights because you’re likely not seeing the full picture and may not focus on the right areas.
It’s not just the speed on the track that matters. If a team doesn’t develop its car, it won’t be able to keep up with other teams who are consistently innovating. And with hundreds of thousands of people watching at the track and millions more tuning in worldwide, there’s no hiding. These models are built and deployed in a matter of weeks; it becomes quickly evident when something goes right or wrong. It’s a uniquely close relationship between what you’re doing and how the fan base reacts, which is part of what makes motorsport so thrilling!
Pairing Human Drivers with AI Technology
There’s a full team of engineers at work, but only two drivers behind the wheel of the cars, with their own driving preferences, sensitivities, and characteristics. The drivers and cars must interact well with each other.
During a race, most viewers see the driver out on the track. They might also get a glimpse of the pit team interacting with the driver during a pit stop, where the car gets refueled, tires get changed, and any other physical adjustments are made. Yet there’s a lot going on behind the scenes that viewers may not know about.
Elsewhere at track, strategists and engineers are watching data come in from the race in real time. They’re using that knowledge to offer insights to the driver once they get back out on the track.
Meanwhile, at the McLaren Technology Centre — which is typically on a different continent from where the race is taking place — more engineers are keeping an eye out for anomalies. Is the car working healthy and doing what it’s expected to do? Or is there potential for it to fail? Identifying those anomalies before they happen and making the proper changes can be the difference between taking the checkered flag and not finishing the race at all.
However, there are too many sensors for humans to look at all the time, so AI serves as an additional engineer. It can identify something it sees, pulling from mountains of data to find a similar situation from the past. Maybe that situation was from a year ago, or perhaps it was only a few races ago. No matter when it occurred, the machine learning model has that prior knowledge and can quickly predict what might happen.
Takeaways for Using Machine Learning
In an ideal world, Formula 1 teams would be able to build and test every possible configuration to see what works and what could use some fine-tuning. They often have limited time and resources, though, which means they must focus their efforts on what will have the biggest performance impact.
Here are a few things the team has learned along the way:
- Data is king. You must treat data with the importance it deserves. It needs to be collected properly, labeled correctly, and implemented the right way or else your results will be skewed. It’s worth spending a little extra effort to ensure there are no issues before running analyses.
- Deployment speed is critical. While properly collecting and labeling data is crucial, you can’t wait for the data to offer perfect results. During the season, if you don’t have a car ready, you miss the race entirely. In other industries, if you don’t deploy a product on time, you miss out on potentially millions of dollars and waste thousands of hours of time. You have to account for uncertainty in deployment, and it’s better to match the need with the tools you have than to sit and wait for perfection.
- It’s okay to fail fast. You might think you have the best solution, but after a few weeks in action, it turns out it’s not delivering the results you need. It’s okay to abandon it — so long as you identify that issue and pivot to something else. Being able to iterate is just as essential as deploying.
There is championship-winning potential in machine learning and AI, but it takes a lot of work to make those massive strides in the space. Instead of taking over our jobs, AI will unlock the potential of our human drivers.
It’s an exciting time to be developing new cars and seeing ideas come to life on track. See you at the next race!