1043 replies to this topic

[Ben1445]

Does each team have a half hour window to set their time? Does that means this is going to take 3 hours? 

[thegamer23]

Allright i tried!!

Hopefully next time around they'll provide better broadcast & the teams will be more prepared for the challenge!

[Beri]

https://youtu.be/8WEtxJ4-sh4

[ExFlagMan]

Well that was exciting!

 

I guess it was the fly on the camera that made the car off - or maybe it was the local taxi service jamming the airwaves.

 

At least the mechanic got to have his lunch.

[maximilian]

I didn't see any of that, but it sounds like an unmitigated disaster... again.

 

Please, just go away.

 

You could literally take ANY off the shelf self driving car that's out there at the moment, and they would do hugely better than this farce, from what I gathered just reading your forum messages.  Not even gonna bother checking the stream recording.

 

Please, just go away.

[HP]

BMW had already a self drifting car a while ago:



Machines are specially good at balancing tasks, so with enough development they can be way better than humans in a racetrack.

 

And that will fuel speculation of cheating  when a human driver does that to perfection in some racing series.

 

I don't mind that these technologies are being developed. A racing series however?

[Ben1445]

Oops. 

 

https://clips.twitch...ldfishEleGiggle

 

https://clips.twitch...lyGoblinVoteNay

Edited by Ben1445, 29 October 2020 - 17:11.

[BRG]

They should really quit while they're behind.

[maximilian]

More embarrassments, I am guessing?  I'm not even bothering to have a look these days.

[ANF]

The first clip was worth the click.
Well, so was the second one.

[Risil]

Beginning to see why we're not all watching self-driving car races yet.

[Jellyfishcake]

Just came in to say I saw the clip of the car driving straight into the wall, absolutely hilarious. 

 

0 interest in this as a series. 

[ANF]

These AI machines will soon outsmart the fastest drivers in the world.

[P123]

Oops. 

 

https://clips.twitch...ldfishEleGiggle

 

https://clips.twitch...lyGoblinVoteNay

 

Machine Learning:  Maldonado Edition

[maximilian]

Beginning to see why we're not all watching self-driving car races yet.

 

The problem isn't so much that self-driving cars wouldn't be able to do it... I am sure there are plenty in development and even use already out there which would do a halfway decent job if slightly reprogrammed for racing.

 

The problem is that Robo"race" is an absolute FARCE.

[krapmeister]

Machine Learning: Maldonado Edition

That's clearly the Grosjean edition.

[ExFlagMan]

Anyone got any idea if this is still a 'living' thing.

 

Found some video from the 20-21 Beta season and it seemed to consist of a single car on track 'racing' against some occasional cartoon character 'ghost' rivals.

[Risil]

Anyone got any idea if this is still a 'living' thing.

 

Found some video from the 20-21 Beta season and it seemed to consist of a single car on track 'racing' against some occasional cartoon character 'ghost' rivals.

 

Roborace appears to be over, and the company that owns it has reassigned its developers to work on its electric van.

 

https://www.autocar....tonomous-racing

 

Looking back at the start of the thread, it's been going on for longer than I remembered. It's sort of mirrored the progress of autonomous driving, which went from various techies and (especially) venture capital hype masters promising fully automated cars were just around the corner, to the realization that at least for now the promise of current technology has been exhausted and it can't deliver the great leap forward.

 

Oh well, we'll still have that meme of the car driving straight into the pit wall at Thruxton. Reposted because I don't think I'll ever stop finding it funny.

 

[ExFlagMan]

Thanks for that.

[midgrid]

That's clearly the Grosjean edition.

 

More like the Katayama edition!

 

[Ben1445]

The Indy Autonomous Challenge did a far better job of putting together a competition for autonomous vehicles.

Just don’t think it was that well noticed outside of the directly relevant industries.

https://en.m.wikiped...omous_Challenge

[pdac]

...

Looking back at the start of the thread, it's been going on for longer than I remembered. It's sort of mirrored the progress of autonomous driving, which went from various techies and (especially) venture capital hype masters promising fully automated cars were just around the corner, to the realization that at least for now the promise of current technology has been exhausted and it can't deliver the great leap forward.

 

 

I've worked in software development for more years than I care to remember and know that when you show someone senior in the company what you've achieved, it's nigh on impossible to get across the fact that 95% of the development is done in the first 5% of the time. They never seem to grasp that if it's 95% working, then it's going to take forever to get that final 5% done (and it's not a usable product until that 5% is finished).

Edited by pdac, 29 July 2022 - 14:59.

[ExFlagMan]

I've worked in software development for more years than I care to remember and know that when you show someone senior in the company what you've achieved, it's nigh on impossible to get across the fact that 95% of the development is done in the first 5% of the time. They never seem to grasp that if it's 95% working, then it's going to take forever to get that final 5% done (and it's not a usable product until that 5% is finished).

 

Doesn't stop them giving the go-ahead to release it though...

[Primo]

Seems a bit ridiculous to do those experiments on full scale cars. Start with typical RC models scale 1:8, on RC tracks, then karts on kart tracks. The tech hurdles will be the same but worked on for a fraction of the cost.

[krapmeister]

Seems a bit ridiculous to do those experiments on full scale cars. Start with typical RC models scale 1:8, on RC tracks, then karts on kart tracks. The tech hurdles will be the same but worked on for a fraction of the cost.

https://www.youtube....h?v=J1DoMRO0ShY

Geez that takes me back - used to race my TC3 touring car there at Hillingdon when I lived in London. I've still got it too! (The car not the skillz...)

[ExFlagMan]

Seems a bit ridiculous to do those experiments on full scale cars. Start with typical RC models scale 1:8, on RC tracks, then karts on kart tracks. The tech hurdles will be the same but worked on for a fraction of the cost.

 

Not sure there is enough room under the bodywork for everything required for autonomous driving - seem to recall the motor, batteries and RC receiver almost filled the space, but it was nearly 45 years ago when I raced them.

[Ben1445]

I suppose you could still do scaled down from 1:1, if not quite as small as a typical RC car as raced there.

[pdac]

I suppose you could still do scaled down from 1:1, if not quite as small as a typical RC car as raced there.

 

But how much would that help and how much cost-saving would there be? If it's minimal, they might as well just go for the full-size version.

[Ben1445]

I mean, an LMP/IndyLights chassis is in the order of $100,000s right?

A kart chassis is more in the order of $1000s, if that.

If you’re trying to get an autonomous racing competition going and to make it as cost effective and accessible as possible, I’d say it’s worth considering.

Not that I have much enthusiasm for the technology anyway, let alone racing it.

[lamo]

But how much would that help and how much cost-saving would there be? If it's minimal, they might as well just go for the full-size version.

Go kart would be good

[Primo]

But how much would that help and how much cost-saving would there be? If it's minimal, they might as well just go for the full-size version.

The cost saving would be huge and the progress rate would most likely be much quicker since repair would be cheap and with that, they can take bigger risks. Tracks would be cheaper to hire, recovery after failures cheaper and faster. 
Maybe a 1:8 is too small, but a 1:5||6 should be able to contain the most important sensors and other tech. 

[prty]

First time that AI beats human champions in drone racing. The same will happen sooner or later with vehicles.

Edited by prty, 02 September 2023 - 07:21.

[ANF]

Probably later.

[flatlandsman]

Where I live, the truck used for roborace is parked up at a company called Arrival, a company that has lost vast sums of money in recent years, I have no doubt the car is in there, I have seen at Upper Heyford doing basic laps of the strip with no-one in it, but I think it is all gone now.  Arrival are very likely to go bust at some point too sadly, but they were very ambitious initially!

[MichaelPM]

First time that AI beats human champions in drone racing. The same will happen sooner or later with vehicles.

 

The residual models part from the 3 minute mark is very interesting. It's human like for the experience drivers gain throughout their careers.

Using a scale car has an issue with human limits are already being surpassed and there is a tendency to ramp up traction limits to gloss over crude programming.

Having the AI using residual models to draw on and compensate for different aero loads or traction limits would be the key to making it interesting to see a full scale car tackle a dynamic track and not just making perfect tyres and aero to show off mechanical limits at break neck speeds.

It's not like track owners or the FIA would be too happy with excessive speeds causing more wear and tear on the track and requiring much stronger safety measures and barriers to contain any accidents.

 

This is a great video with F1 type leaps in development of software and hardware to push the limits.

[maximilian]

No doubt that machines CAN do this, but "Roborace" will still go down in history as a ludicrous failure that was embarrassing and painful to watch, especially after how they hyped it up initially.

[pdac]

It's a case of trying to implement what's possible rather than what's achievable. I work in software and we often tell people that producing a demonstration requires 5% of the total effort; there's 95% more to do to produce a finished product.

[Grippy]

186mph Self-Driving AI Cars Will Soon Race on F1 Tracks

https://greekreporte...ing-ai-cars-f1/

 

extract  ".....On April 28, 2024, the Yas Marina circuit will host a competition with a prize of $2.25 million (£1.85 million) on the line. What makes this event truly unique is that instead of traditional racing teams with engineers, drivers, and mechanics, the competing teams will exclusively consist of members from universities and research teams specializing in artificial intelligence (AI).

 

Participating in this competition are teams representing institutions from various parts of the world, including North America, Italy, Abu Dhabi, Singapore, Germany, and Hungary, reported BBC Science Focus.

The vehicle specifications

At the GITEX event, the specifications of the self-driving cars will remain consistent for all competing teams. The vehicle weighs 690 kilograms with a maximum speed of 186mph. It is powered by electricity, which places it in a similar category to Formula E, the electric racing series.

Viewers will have the opportunity to access the race through virtual and augmented reality experiences. They can virtually sit in the driver’s seat and fully take the experience of the race, according to BBC Science Focus......"

 

That appeals to my inner geek.

[AncientLurker]

Sounds like a lot of fun. I hope it is all cars on track together and not time attack. That said, if they all race together I would be making a robust car capable of a few withstanding a few licks.

[GreenMachine]

T/A, lap dash, sprint - all fine by me, as long as track limits strictly policed.  Just going fast(est) on a racing circuit without hitting things would be a great start, and an improvement on their predecessors   

 

W2W collision avoidance with other competitors would be fairly simple I imagine (relatively speaking) except for the imperative to win - balancing those two requirements could be a challenge I reckon!

[Grippy]

Yeah I really want to see how they manage. There have been a lot of developments with LIDAR, computing power and AI, with various robo-taxis, delivery pod things and drones, so it will be interesting how they compute racing lines, track limits and other vehicles doing the same (assuming it's not time attack).

[YamahaV10]

What I'd like to see is Indycars back on super speedways. Have them remote controlled by drivers at sim rig stations from their infield trailer.

 

Crank them up to 1200 horsepower and allow them to use as much or little downforce as they want. Even Talladega could be used. Talladega would be the fastest race of them all.

 

The fans would still be at risk but maybe they could just be televised races. Or extra fencing reinforcement would have to be used. Or grand stands close to the track would have to be sectioned off.

 

With driver safety a non factor, it would be like the good ol days again. Engineers could literally do anything they could to make the cars go around the track faster. Plus the cars would be a fair bit lighter. 

[GreenMachine]

What I'd like to see is Indycars back on super speedways. Have them remote controlled by drivers at sim rig stations from their infield trailer.

 

Crank them up to 1200 horsepower and allow them to use as much or little downforce as they want. Even Talladega could be used. Talladega would be the fastest race of them all.

 

The fans would still be at risk but maybe they could just be televised races. Or extra fencing reinforcement would have to be used. Or grand stands close to the track would have to be sectioned off.

 

With driver safety a non factor, it would be like the good ol days again. Engineers could literally do anything they could to make the cars go around the track faster. Plus the cars would be a fair bit lighter. 

 

I'll watch that on TV thanks!

 

At least making them lighter would limit the results when MV² comes into play ... 

[maximilian]

What I'd like to see is Indycars back on super speedways. Have them remote controlled by drivers at sim rig stations from their infield trailer.

 

Crank them up to 1200 horsepower and allow them to use as much or little downforce as they want. Even Talladega could be used. Talladega would be the fastest race of them all.

 

The fans would still be at risk but maybe they could just be televised races. Or extra fencing reinforcement would have to be used. Or grand stands close to the track would have to be sectioned off.

 

With driver safety a non factor, it would be like the good ol days again. Engineers could literally do anything they could to make the cars go around the track faster. Plus the cars would be a fair bit lighter. 

 

I would watch that, too!  Bring it!