Are F1 team using AI to gain advantage?

[CHT]

With the fast growing AI machine capabilities and applications, I am just wondering if F1 teams are using AI in area such as race strategies or engine cooling design etc.

In chess therr is AlphaZero and I am sure AI may have the capability to help in race strategies

[CHT]

https://techcrunch.com/2024/04/14/how-n ... rmula-one/

Williams is using NSC AI for aero design.

[Hoffman900]

https://techcrunch.com/2024/04/14/how-n ... rmula-one/

Williams is using NSC AI for aero design.

None of this stuff is actually AI though, if we define AI as actually being intelligent.

You have been able to do similiar in 1D engine gas pressure dynamics programs for over a decade now.

[Greg Locock]

A rather less breathless article https://www.neuralconcept.com/post/form ... rrelations that seems to be missing some rather significant components

So presumably the idea is you test millions of CFD models that aren't actually a team's models, and use the insights from those to comment on the proposed design.

[moro.leonaers]

We are not there yet on the track, the human element is KING. Imagine AI algorithm that is able to calculate the time it would take to clear the car and calculate when the virtual safetycar would be called off to a green flag while the marshall were trying to push the car off the track...

AI could be used if enough data was available to simulate strategies in case of car system failures, but at the end of the day the driver drives, track conditions and tire deg are pretty much never a constant variable.

I think AI can be used to gain a very small strategic advantage in the developement of the car through the year tho. which they are probably using in some shape or form. But its such a complex topic.

[Hoffman900]

A rather less breathless article https://www.neuralconcept.com/post/form ... rrelations that seems to be missing some rather significant components

So presumably the idea is you test millions of CFD models that aren't actually a team's models, and use the insights from those to comment on the proposed design.

Assuming you have good correlation, which in 2024 is still a struggle for all industries in which CFD is used, not just F1.

[Zynerji]

I stated this years ago. I believed Mercedes had a better iterative design loop AI than anyone during their run.

When the hardware catches up to run full scale point cloud models with simultaneous CFD, FEA and torsional dynamics(cymatic resonance), the refinement algorithms will be King, and forward generational.

F1 teams are about to be 150ppl again.

[Seanspeed]

I stated this years ago. I believed Mercedes had a better iterative design loop AI than anyone during their run.

When the hardware catches up to run full scale point cloud models with simultaneous CFD, FEA and torsional dynamics(cymatic resonance), the refinement algorithms will be King, and forward generational.

F1 teams are about to be 150ppl again.

I dont know about AI, but Mercedes probably has benefited from the close partnership with AMD, who have been making not only the best CPU's for HPC(high performance compute), but also top tier GPU accelerators for high precision compute(FP64) which is exactly the combination you want for the best CFD performance. I have no doubt Mercedes invited AMD software folks to help optimize their simulation software around their hardware as well to really maximize their results.

[CHT]

Modern F1 is a very database driven sports which requires speedy data processing to make quick decision on and off track and an AI machine will certainly be able to digest information much quicker than human mind, and run through millions of possible simulations and comparison within very short time before deciding which are the best possible outcomes etc.

AI is a self-learning machine, which mean the more data it analyze the smarter it will get, If supposed all historical race data such as track temperature, winds, tyre compound, track position etc put into good use, AI machine are more likely to be able to make suggestion which human may missed out. And perhaps also the same with CFD when designing a completely new car for new regulations. And certainly it will be cheaper too because team will use less people, and hence benefit in budget cap.

[Greg Locock]

" run through millions of possible simulations and comparison within very short time" turn round, you sound a bit muffled.

[PlatinumZealot]

With the fast growing AI machine capabilities and applications, I am just wondering if F1 teams are using AI in area such as race strategies or engine cooling design etc.

In chess therr is AlphaZero and I am sure AI may have the capability to help in race strategies

Well they have always been using machine learning to get the best setups for the suspension, aero and Hybrid system... It's just that AI is snazzy new buzz word now.

[Greg Locock]

Once upon a time in 1847 there was gradient descent, as one of the first formal optimisation methods. During one of the AI winters AI coopted the entire field of optimisation as part of AI, giving it an increased funding stream and the ability to actually discuss something useful in their papers. Or so it seems to me.

[CHT]

With the fast growing AI machine capabilities and applications, I am just wondering if F1 teams are using AI in area such as race strategies or engine cooling design etc.

In chess therr is AlphaZero and I am sure AI may have the capability to help in race strategies

Well they have always been using machine learning to get the best setups for the suspension, aero and Hybrid system... It's just that AI is snazzy new buzz word now.

Computing data processing is different from generative AI I think.

[PlatinumZealot]

With the fast growing AI machine capabilities and applications, I am just wondering if F1 teams are using AI in area such as race strategies or engine cooling design etc.

In chess therr is AlphaZero and I am sure AI may have the capability to help in race strategies

Well they have always been using machine learning to get the best setups for the suspension, aero and Hybrid system... It's just that AI is snazzy new buzz word now.

Computing data processing is different from generative AI I think.

Generative AI uses neural networks to "generate new" from seemingly random, extremely large or disorderly data sets. The Data in Formula 1 is not as random or disorderly as you think, and the problems to solve are not unique, or individualized. Most teams have the same problems. That said, Generative AI is still applicable. Lets look at your question in two parts. Strategy and then Car design.

Historical data on tracks, weather conditions, past scenarios with safety cars, live decisions by other teams are large data sets. Ultimate the team wants to know when to pit for tyres. Generating new scenarios from data sets can also be used to create simulations. The AI can even look at driver inputs and suggest ways for the driver to brake earlier, steer accelerate etc. We KNOW that teams do this because this was mentioned by Toto Wolff back in 2016? When Hamilton set the pole in Singpaore he was said to surpass all the different simulations. Was it called generative AI back then? Hmm maybe..

SO HOW is this different from Machine learning you ask?

Generative AI is for generating NEW data based on data is exposed to, Machine learning is a wider set of techniques for creating the models to analayse the data to make decisions and predict on that data.

So onto the car design . Machine learning is what you would use for these very "constrained" and "guided" problems (regulations!) of a physical nature to help you make decisions. You could generate a fancy rear wing if you want with generative AI but that might be very out there, you have to test if it it works and you might have to test hundreds of these! but you can use machine learning to make the model of how to design a rear wing, and then choose the best design.


This is Chat GPT answers on the two in Formula 1:

In the context of optimizing race strategies in Formula 1, machine learning techniques such as reinforcement learning may be used to develop adaptive strategies, while generative AI techniques may be less directly applicable.

[CHT]

Exploring his areas in more depth, AI can certainly help with understanding and planning of race strategy and give the Red Bull pit wall the edge when it comes to attacking races.
Computers could also offer guidance on the setup choice for a race weekend, and also even help lead the car designers on the directions that would be best placed to take when it comes to upgrade developments.

Plus, do not forget the impact of the cost cap this year which means teams are needing to put efficiency above almost everything else when it comes to deciding where they deploy resources.
Horner adds: “Making the best decision you can to develop your car cost efficiently, so cost-effective performance, is absolutely crucial for us as we move forward with the lack of testing.
https://www.motorsport.com/f1/news/red- ... p/6271315/


If RBR is to input all all AN's design concept, philosophy, rules and limitation of into their AI machine, I am pretty sure the AI will be able to do a pretty decent job in help to generate ideas and solutions.