Post here all non technical related topics about Formula One. This includes race results, discussions, testing analysis etc. TV coverage and other personal questions should be in Off topic chat.
Is it being used already?
The reason for asking is that with the passing of time, so much accumulated data can be used to solve conundrums maybe better than is being done presently.
Is it being used already?
The reason for asking is that with the passing of time, so much accumulated data can be used to solve conundrums maybe better than is being done presently.
They already are as a very cursory google search shows.
Is it being used already?
The reason for asking is that with the passing of time, so much accumulated data can be used to solve conundrums maybe better than is being done presently.
They already are as a very cursory google search shows.
Thank you for that.
its just that it hadnt seen discussed in here before, so I jumped to the wrong conclusion.
From reading those articles it reveals the tip of the iceberg,
I imagined a couple of areas it could be used is being used is driver or rider selection and avoiding the Masi incident
Etc
ML in particular is widely used in vehicle dynamics, as the first article says, the fundamental issue is making sense of vast amounts of data. Bear in mind that the naughty kiddies in the AI world have co-opted multidimensional optimisation as part of AI, so when you use gradient descent then you are using AI. Not exactly HAL9000.
It really depends on what your goal is, in F1 are they trying to break down a lot of data and derive intelligent outcomes from that? I don't see why they would use this track side, but could see a very good use case for it when using telemetry post race and in testing scenarios.
AWS which provides some of the stats on overtake likelihood % and laps till overtake etc uses machine learning to provide some of this data.
It will no doubt become more prevalent as it develops but a common misconception is just how powerful a machine is needed to produce decent machine learning. The AWS data is nice but not revolutionary and it is utilising clusters of powerful machines. AI and machine learning for the common man and even monied business is still very early stages and the term "Machine learning" is often used with some very basic algorithms. This system would have to be able to evaluate and ever changing situation for 20 cars for which the data can pivot rapidly inf there's an accident or weather change. It would have to be able to collect a lot of data for all 20 cars and react with outputs within seconds.
What may happen first is the ability to use machine learning to understand correlation issues between track and CFD and to produce reliable mathematical adjustments for the CFD where necessary as this is something that could be done with enough data.
But a lot can change in a short space of time and F1 has a lot of money to spend on this kind of software.
What you can take away is that is inevitable that Ai and Machine Learning could revolutionise the sport within ten years, and that is conservative. But would it be banned by the FIA in terms of development and strategy? I hope so, I want sport to be human, not about the best software developers.
What may happen first is the ability to use machine learning to understand correlation issues between track and CFD and to produce reliable mathematical adjustments for the CFD where necessary as this is something that could be done with enough data.
We already do that (not CFD but dynamics and K&C), so the F1 boys do at least know about it. We call it auto-correlation, a very very poor choice of phrase since that is already one specific thing.
We have a thread on this already actually.(I think when Amazon AWS came into F1)
The strategis at force india was asked this question and she says that it is not really used only in small basic applications.
The original point of the topic was what is said in the title as an example of it's usage.
The fact that there are articles out there saying they use it doesnt show even a small detail of how it works and the more importantly examples of significant results, and as distinct from correlation between data and simulations.
The AWS stuff is fairly straightforward from what my very limited knowledge understands, but I was imagining much more complex usage, which by the way when I started considering this subject because of the Masi incident, is limited to the singular systems of PLC logics, iterations in basic CFD simulations, differential diagnosis, and plain old what if's (in spreadsheets or otherwise) and some very basic correlating between engine data and simulations.
So having tried to focus the subject a bit more this lecture from Yale University puts some knowledge and perspective on the subject in a broader sense.
signed
busted arse ex race mechanic
The original point of the topic was what is said in the title as an example of it's usage.
The fact that there are articles out there saying they use it doesnt show even a small detail of how it works and the more importantly examples of significant results, and as distinct from correlation between data and simulations.
The AWS stuff is fairly straightforward from what my very limited knowledge understands, but I was imagining much more complex usage, which by the way when I started considering this subject because of the Masi incident, is limited to the singular systems of PLC logics, iterations in basic CFD simulations, differential diagnosis, and plain old what if's (in spreadsheets or otherwise) and some very basic correlating between engine data and simulations.
So having tried to focus the subject a bit more this lecture from Yale University puts some knowledge and perspective on the subject in a broader sense.
signed
busted arse ex race mechanic
Interesting what you said about PLC... Yes, PLC is more of switching and controlling type of purpose. I think this is too basic and too "sudden" for any complicated incident so that is why Masi and Co. still have to deliberate over things. In the example a damaged barrier, a marshall has to inspect then send a signal to race control, then race control can decide if double waved or red-flag based on the situation at that location and other locations. If it were 100% down to the PLC programme would would need in the first place have to have these other situations on their own progamme ladders, and let's say it did, it would throw flags extremely fast without any form of judgement. So yeah, PLC should be limited to just switching and controlling.
The "meatier" judgement and control through machine learning, similar to what companies like Tesla, ZooX (Amazon) and Goolge does would be interesting to see for race control, but I know because human lives are at stake, they would have have it as only a suggestive sort of thing. Applying this for strategies I think as you said, is perfect application of AI and machine learning. But it would sorta lead us to the "efficient choices" situation: if all teams made the most efficient choices, then, will strategy not matter anymore?
I'm taking a liberty here, but thought that this example of another high level system of crisis management would be of interest.
The idea of this thread was to look at a similar approach where what if scenarios had a calculated solution rather than arbitary human deficit decisions like at Abu Dhabi 2021 debacle.
And surely for this particular pitot problem they could engineer a wire mesh cover??
Raising this topic was initially brought about by the kerfuffle of Abu Dhabi 2021 with its enormous and ongoing consequences. Motor racing being what it is throws up so many scenarios some of which are new and some repeats, overlay that with human nature and the rubbery thing called rules and we have less than optimal outcomes.
Could this method provide better and more consistent circumstances.
F1 could throw a bob to someone like MS (they need it ) to see if it could help. It would be worth it because of the high stakes.
) to see if it could help. It would be worth it because of the high stakes.