Possibilities for Machine Learning

[PlatinumZealot]

Machine learning can solve it quickly if they use machine learning for that.

[dialtone]

Machine learning can solve it quickly if they use machine learning for that.

Machine learning is not a magic bullet like that, you need to have the problem you want to solve be extremely specific and precisely measured.

[PlatinumZealot]

Machine learning can solve it quickly if they use machine learning for that.

Machine learning is not a magic bullet like that, you need to have the problem you want to solve be extremely specific and precisely measured.

The context was seven post rig and data from the track to modify the suspension behaviour.
Should get results quickly if it were a suspension problem.

So i feel it's an aero issue.

[dialtone]

Machine learning can solve it quickly if they use machine learning for that.

Machine learning is not a magic bullet like that, you need to have the problem you want to solve be extremely specific and precisely measured.

The context was seven post rig and data from the track to modify the suspension behaviour.
Should get results quickly if it were a suspension problem.

So i feel it's an aero issue.

data for ML is just to train the model, but you still need to know exactly what you are looking to optimize with the design. Not saying you are wrong, but ML is not magic.

EDIT: in particular ML can't work if you have correlation issues, at best it would be something like a GAN that attempts to design various shapes in an attempt to minimize a few behaviors but you really need to know that the synthetic tests you run in your design app are going to behave like real life or it's useless.

[PlatinumZealot]

So you don't think Mercedes has been taking precise measurements of the suspension and the airflow over the car?

I believe they have enough data to input into a machine learning model.

The concept is viable. Here some researcher's use it to improve a suspension conrtol system based on different bumps in the road:

https://www.extrica.com/article/22025

Very much can be applied to F1 suspension design and tuning.

[PlatinumZealot]

Machine learning is not a magic bullet like that, you need to have the problem you want to solve be extremely specific and precisely measured.

The context was seven post rig and data from the track to modify the suspension behaviour.
Should get results quickly if it were a suspension problem.

So i feel it's an aero issue.

data for ML is just to train the model, but you still need to know exactly what you are looking to optimize with the design. Not saying you are wrong, but ML is not magic.

EDIT: in particular ML can't work if you have correlation issues, at best it would be something like a GAN that attempts to design various shapes in an attempt to minimize a few behaviors but you really need to know that the synthetic tests you run in your design app are going to behave like real life or it's useless.

Which shapes are you refering to here?

[S D]

So you don't think Mercedes has been taking precise measurements of the suspension and the airflow over the car?

I believe they have enough data to input into a machine learning model.

The concept is viable. Here some researcher's use it to improve a suspension conrtol system based on different bumps in the road:

https://www.extrica.com/article/22025

Very much can be applied to F1 suspension design and tuning.

I'm sure that they can learn a lot from machine learning for suspension but in order to improve over a race weekend, they need to factor in which tire is used, tire wear, tire temperature, adjustments on FW after each tire change, tracks rubbering in, fuel loads changing with each lap, then ride height changes with dropping fuel load, track temperature changes as the sun goes down, following a car or in open air, DRS on and off, etc etc etc.

[dialtone]

The context was seven post rig and data from the track to modify the suspension behaviour.
Should get results quickly if it were a suspension problem.

So i feel it's an aero issue.

data for ML is just to train the model, but you still need to know exactly what you are looking to optimize with the design. Not saying you are wrong, but ML is not magic.

EDIT: in particular ML can't work if you have correlation issues, at best it would be something like a GAN that attempts to design various shapes in an attempt to minimize a few behaviors but you really need to know that the synthetic tests you run in your design app are going to behave like real life or it's useless.

Which shapes are you refering to here?

Any shape they have measurements on and have a decent behavioral model about, could be a wing, a suspension and so on.

If you have good correlation, the data and a nice way to model and simulate the end result, you can actually design parts via machine learning because it speeds up the feedback loop, but if you don't have even one of the above...

[matteosc]

I thought we were joking about machine learning... They need to understand the problem and figure out how to fix it: there is no machine learning that is going to do that for them.

[wesley123]

So you don't think Mercedes has been taking precise measurements of the suspension and the airflow over the car?

I believe they have enough data to input into a machine learning model.

The concept is viable. Here some researcher's use it to improve a suspension conrtol system based on different bumps in the road:

https://www.extrica.com/article/22025

Very much can be applied to F1 suspension design and tuning.

I don’t think anyone is doubting that ML can be useful, but an AI needs to know what it should learn, which is particularly difficult if you are lost. And just like needing to know what to learn, you need to know what you are looking to achieve.

[matteosc]

So you don't think Mercedes has been taking precise measurements of the suspension and the airflow over the car?

I believe they have enough data to input into a machine learning model.

The concept is viable. Here some researcher's use it to improve a suspension conrtol system based on different bumps in the road:

https://www.extrica.com/article/22025

Very much can be applied to F1 suspension design and tuning.

I don’t think anyone is doubting that ML can be useful, but an AI needs to know what it should learn, which is particularly difficult if you are lost. And just like needing to know what to learn, you need to know what you are looking to achieve.

Agree. ML needs a set of inputs and outputs to be "trained". If you do not have the data you do not go anywhere. Also this is not the kind of problem you want to feed to a AI.

[PlatinumZealot]

data for ML is just to train the model, but you still need to know exactly what you are looking to optimize with the design. Not saying you are wrong, but ML is not magic.

EDIT: in particular ML can't work if you have correlation issues, at best it would be something like a GAN that attempts to design various shapes in an attempt to minimize a few behaviors but you really need to know that the synthetic tests you run in your design app are going to behave like real life or it's useless.

Which shapes are you refering to here?

Any shape they have measurements on and have a decent behavioral model about, could be a wing, a suspension and so on.

If you have good correlation, the data and a nice way to model and simulate the end result, you can actually design parts via machine learning because it speeds up the feedback loop, but if you don't have even one of the above...

Just suspension control we were talking about. You can isolates the forces on the suspension and emulate the aero forces on a suspension rig so the shape of the wing doesn't come into play.

All of this was a response to my guy who said that the suspension alone can solve the issue, so I gave him scenario of how the teams could solve that problem if suspension alone was the issue b

[PlatinumZealot]

So you don't think Mercedes has been taking precise measurements of the suspension and the airflow over the car?

I believe they have enough data to input into a machine learning model.

The concept is viable. Here some researcher's use it to improve a suspension conrtol system based on different bumps in the road:

https://www.extrica.com/article/22025

Very much can be applied to F1 suspension design and tuning.

I don’t think anyone is doubting that ML can be useful, but an AI needs to know what it should learn, which is particularly difficult if you are lost. And just like needing to know what to learn, you need to know what you are looking to achieve.

Agree. ML needs a set of inputs and outputs to be "trained". If you do not have the data you do not go anywhere. Also this is not the kind of problem you want to feed to a AI.

So Mercedes don't have any data for a machine learning model after two months and 4 races. Hmm OK.

Let's just move on then to other developments.

[dialtone]

I don’t think anyone is doubting that ML can be useful, but an AI needs to know what it should learn, which is particularly difficult if you are lost. And just like needing to know what to learn, you need to know what you are looking to achieve.

Agree. ML needs a set of inputs and outputs to be "trained". If you do not have the data you do not go anywhere. Also this is not the kind of problem you want to feed to a AI.

So Mercedes don't have any data for a machine learning model after two months and 4 races. Hmm OK.

Let's just move on then to other developments.

I think we're bordering on the OT here but... Driving 60 laps with the exact same setup is useless for ML. When you build a model, you need to setup the car in many different ways and then collect enough data on each setup so that, whatever computer model you have, it can lookup the behavior of the system at a given setup.

Once you have enough setups collected, and metrics on each of them, you should end up with a decent range of parameters to be able to run regressions on the data and generalize the model of how your suspension works.

At this point, even with all this data and model, you still need to know what you want to accomplish as you can't just type: "how to solve porpoising" in the computer and it will figure it out. First Merc needs to find the trigger of said behavior or they need to come up with the exact behavior that they want the suspension to work, and actually know (otherwise it's a bet) that it will solve porpoising.

Once you have the final model, and all of the setup plus its data, only then you have barely enough information to digitally make a new suspension, which then needs to be validated for manufacturing, weight, size and so on. And last you can test it in the car and if that doesn't work you are free to figure out what went wrong in any of the previous steps.

And keep in mind that anyway the computer needs to know what are its degrees of freedom when designing the suspension anyway, so if whoever is setting the experiment up misses a key freedom in the design, but chooses to constrain it in a particular way, you will end up with no possibility of a design that solves your issue.

All in all I doubt any team is actually doing this, what's more likely is instead that the ML system is going to tell you what is the best setup given everything that you have, as is. ML aided design is pretty limited for this stuff IMHO.

[matteosc]

I don’t think anyone is doubting that ML can be useful, but an AI needs to know what it should learn, which is particularly difficult if you are lost. And just like needing to know what to learn, you need to know what you are looking to achieve.

Agree. ML needs a set of inputs and outputs to be "trained". If you do not have the data you do not go anywhere. Also this is not the kind of problem you want to feed to a AI.

So Mercedes don't have any data for a machine learning model after two months and 4 races. Hmm OK.

Let's just move on then to other developments.

I am not sure that you fully understand which kind of data you would need to find a solution to the porpoising via ML.
Most people see Artificial Intelligence as a magic trick to find solutions to any kind of problems, when the reality is way more complex and somehow "limited". But on the other hand you may be a AI expert, so my apologies if that is the case.