RoboRace - An autonomous support series for Formula EOpen Wheel

[Pupik]

Jokes aside, I'm kind of interested to see how a race works out. This is a logical evolution of autonomous vehicles, and seeing them handle limits to which humans may not be able to achieve might have serious use for the efficient and safe future of the automobile (and its passengers/cargo).

But I can't see myself interested in a series of races, unless the events can really offer something that other championships do not, because they're restricted by human limits, artificial engineering rules, or restrained by historical factors. I feel one way to wow the masses would be by having these conveyances handle varying terrain.

 

[Liquid]

If you fast forward to the future, and autonomous racing has advanced sufficiently enough for something like an Adrian Newey X2010 to be a reality, you'd have to have dedicated racetracks for these series. In a way, current racetracks are built according to perceived human reflexes in racing. There's no way you could have spectators at current tracks with cornering speeds of 150mph on what might usually be a 60mph corner. If that went wrong, it would go badly wrong.

 

[TenEightyOne]

There is no need for flags at all.

There is if you want it to be a spectator sport - viewers need to know what's happening. Marshalls might also like some visible signal that they're protected while disentangling M.Rossi and N.Mansell.

 

[Dylan]

Watching a human driving a car to the best of their ability vs. robot cars that are programmed to follow the same racing line each lap? I'll stay with human drivers. That's just my opinion.

 

[MatskiMonk]

There is if you want it to be a spectator sport - viewers need to know what's happening. Marshalls might also like some visible signal that they're protected while disentangling M.Rossi and N.Mansell.

For its primary purpose I think clinging to a flag based system is a bit redundant, though I guess there are secondary elements that it does address too.

 

[TenEightyOne]

Watching a human driving a car to the best of their ability vs. robot cars that are programmed to follow the same racing line each lap? I'll stay with human drivers.

Definitely the right attitude. However, I think the sport described in the OP isn't likely to be anything like that - what would be the point?

 

[Bo]

I'm finding the "let's watch cars drone round at the same speed and form a procession" comments rather humorous.

Robotics is not an endeavor in line-following. It may have been 40 years ago, but technology is at the point where robotics equates to programming things to mimic the thought process of a living being. Robots may not be able to think for themselves, but we can certainly make them act like they can by programming them that way. If any of you have ever watched robot sports before, they're wonderfully entertaining - it's a real test of technological skill. The cars might be a bit jerky or slow the first time we see them, but they will be programmed to understand a track or a race scenario like any human would do. In time, they will try and use different parts of the track, they will try and out-brake each other, and they will definitely be aiming to make overtaking moves.

It's interesting to see that Toto Wolff has already refused to even acknowledge the concept as a sport, but I wouldn't be turning to the F1 paddock to seek balanced and objective views of series outside of F1.

 

[B.K]

I thought it couldnt get any more stupid than Formula E already is. But I was wrong.

 

[05XR8]

The flag system will most likely be remote commands rather than visual cues. So, let's say for the yellow flag


A very basic (and no doubt wrong) code where the variable (flags.yellow) is increased up to 1** then the speed will be from 0-30 mph depending on track speed data because if it was just 30, they wouldn't brake for a corner if they needed to.

**thinking about it, it's more likely to be greater or equal to 1 for different zones in which case they'd have to add different variables for different sectors

So, if a car does go down a runoff area in street circuit, cars flip or electrical failures, humans will still have to push them out of the way or reset them or lift them. Like slot cars.

Watching a human driving a car to the best of their ability vs. robot cars that are programmed to follow the same racing line each lap? I'll stay with human drivers. That's just my opinion.

That's Monaco in a nut shell.

On the other hand, if these cars are programmed using AI from PDI count me out. AI from T10 in Unbeatable setting, a bit better.

 

[Dylan]

That's Monaco in a nut shell.

 

[TenEightyOne]

So, if a car does go down a runoff area in street circuit, cars flip or electrical failures, humans will still have to push them out of the way or reset them or lift them.

I think you're missing the point - he means that flags would be un-necessary because course safety data would be transmitted directly to each racer's CPU, not because there isn't a need for actual safety calls.

 

[Downhill Dino]

Thinking about it, it'll probably look no different then if I were to load up almost any racing video game and watch the AI run around.

 

[VXR]

Thinking about it, it'll probably look no different then if I were to load up almost any racing video game and watch the AI run around.

Just what I was thinking. I and millions of others have already participated in this during GT5 b-spec.

 

[Bo]

I want PD vs SMS.

One team's cars can brake early and lift off for no apparent reason, the other team's cars can divebomb at will and run everyone else off the road.

 

[TeamCZRRacing]

I can't wait for Team Polyphony Digital.

I can't wait for Slightly Mad Motorsports, so they can design a car that drives erratically and knocks every other car within a 500-foot radius off the track.

EDIT: Somewhat treed by @Bo

 

[Northstar]

I'm finding the "let's watch cars drone round at the same speed and form a procession" comments rather humorous.

Robotics is not an endeavor in line-following. It may have been 40 years ago, but technology is at the point where robotics equates to programming things to mimic the thought process of a living being.

Won't a robot be able to figure out the optimum line much faster than a human though? It seems like a procession would be even more inevitable than an F1 race since the A.I. would be able to execute at optimum speed at a faster and more consistent pace than a human.

I'm not going to completely write it off until it's actually completed a few races. But I just can't see it being all that interesting unless they don't have any practice or qualifying sessions.

 

[Bo]

Won't a robot be able to figure out the optimum line much faster than a human though? It seems like a procession would be even more inevitable than an F1 race since the A.I. would be able to execute at optimum speed at a faster and more consistent pace than a human.

Well, theoretically, but that depends completely on how well programmed each car is. If the track was a perfectly flat sandbox it would end up being pretty straightforward, but they have to factor in bumps, grip changes, a lot of things that drivers adapt to on "feel" alone - since a computer can't "feel" and think independently how to react, having the optimum line could be fast, slow, or ending up with the car in the wall.

I think people are going to have to re-learn their definitions of motorsport when this comes along. The racing in the end will be a very different product to manned motorsport, and the races will offer a very, very odd sight. The first few races will probably look like a mess - we'll have crashes, confused cars, electrical issues and lots of random or delayed movements. I think a few people are expecting slightly too much of the cars from the get-go.

I'm interested in what the cars will look like, though. Racing cars, and all vehicles, have always been designed around a human being. No survival cell or cockpit will create some very intriguing LCG designs.

 

[Downhill Dino]

Autonomous cars in the past know how to move about a race track, not sure why you seem to doubt their abilities in comparison to humans. Again, it'll be much the same as video game AI.

 

[Bo]

Simply moving around a track is a lot, lot different to direct racing.

 

[Downhill Dino]

What do you mean 'direct racing'?

 

[Furinkazen]

This is a techheads dream, a petrolheads nightmare. I'm going to be following this very closely.

 

[Bo]

What do you mean 'direct racing'?

Cars in close proximity trying to beat each other?

All current autonomous cars are able to drive around tracks because they do so at a safe speed. They only have to learn one line and drive it at a safe pace. This is the opposite of that - the cars are trying to race, so they'll be trying to go faster wherever they believe they can. They won't follow one line, because they'll understand that the objective of a race is to finish ahead of every other car, so they will have to take calculated risks based on their surroundings.

The cars won't be completely blind. Autonomous cars work by using a number of sensors to gather their basic surroundings - in this case, where the walls are, where the apex is, where other cars are around it. What they do with that information then depends upon how they're programmed.

I'm only basing this off existing robot sports. They're pretty weird to watch, but they really are entertaining.

 

[MatskiMonk]

Autonomous cars in the past know how to move about a race track, not sure why you seem to doubt their abilities in comparison to humans. Again, it'll be much the same as video game AI.

I'm not so sure it will be the same as videogame AI, AI for racing games is programmed to simulate human drivers, rather than get round the track in the most methodical and efficient way.

 

[LeGeNd-1]

I'm quite torn on this concept. On one hand it would help accelerate the development of autonomous transport tech and allow us to see racing cars designed beyond human limits. But at the same time it will be a case of who have the best software/brains at the factory. People are already complaining as is when one team dominates (it's all in the car, not enough focus on the driver, etc), it's gonna be worse with new emerging technology.

Rather than a fully autonomous car, I'd rather see a remote controlled car. It's like a crossover between sim racing and real racing. The drivers use VR and steering controllers, and the inputs are beamed to actual cars on track. The feeds from on board camera are beamed to the drivers back at the garage. You still get the element of human competition, but you remove the danger aspect from the track and are free to design cars as fast as possible.

If it does ends up being fully autonomous, I'd love to see a car vs bike race, Project Gotham style, with this:

 

[TenEightyOne]

But at the same time it will be a case of who have the best software/brains at the factory.

If you look at a Formula series like F1 you could say that that's what happened. They all had the same design parameters/limitations but IRBR-Renault and Mercedes were worlds apart.

Rather than a fully autonomous car, I'd rather see a remote controlled car. It's like a crossover between sim racing and real racing. The drivers use VR and steering controllers, and the inputs are beamed to actual cars on track.

RC racing already exists, I think, if not with the VR aspect.

You still get the element of human competition, but you remove the danger aspect from the track and are free to design cars as fast as possible.

I think there's a potential for greater danger; at the moment most drivers will pull out of a dangerous move based on self-preservation (insert Maldonado joke here). We can program AI systems to attempt self-preservation but we can't make them instinctively understand it.

And what about the crowds? If we went to the zenith of development and saw cars pulling 20 lateral Gs where would we put the crowds? There are futuristic solutions to crowd placement but they're many many years off. Right now there are no tracks that could support the kind of racing you describe.

 

[LeGeNd-1]

If you look at a Formula series like F1 you could say that that's what happened. They all had the same design parameters/limitations but IRBR-Renault and Mercedes were worlds apart.

That's exactly what I meant. The gulf in performance is gonna be so large one team that manage to figure out the best AI algorithm will just run away with it. It's going to be more a tech demo than a race.

RC racing already exists, I think, if not with the VR aspect.

I know RC racing exists, but has there been one with full scale cars?

I think there's a potential for greater danger; at the moment most drivers will pull out of a dangerous move based on self-preservation (insert Maldonado joke here). We can program AI systems to attempt self-preservation but we can't make them instinctively understand it.

And what about the crowds? If we went to the zenith of development and saw cars pulling 20 lateral Gs where would we put the crowds? There are futuristic solutions to crowd placement but they're many many years off. Right now there are no tracks that could support the kind of racing you describe.

I think you can program self preservation by tying it to risk reward benefit. Obviously if you crash you lose the race. I think we'll see more sensible passes by fully autonomous cars actually.

Yeah crowd control is going to be a problem without completely modifying existing track barriers and grandstands. I dunno, maybe limit on track spectators only to enclosed buildings. Or set up large viewing areas far from the track (kinda like old drive-in theatres).

 

[TenEightyOne]

That's exactly what I meant. The gulf in performance is gonna be so large one team that manage to figure out the best AI algorithm will just run away with it. It's going to be more a tech demo than a race.

But they still have to have a better car too - to continue the F1 parallel look at Ricciardo in this year's IRBR-Renault. I think he's WDC material but he was scuttering around the midfield for most of the year.

Yeah crowd control is going to be a problem without completely modifying existing track barriers and grandstands. I dunno, maybe limit on track spectators only to enclosed buildings. Or set up large viewing areas far from the track (kinda like old drve-in theatres).

You have to be making a looooot of money from races to even consider that. I'm willing to bet it's not going to happen. By the time AI racing is a big thing (if it ever is) then we'll be watching it trackside in VR.

 

[LeGeNd-1]

But they still have to have a better car too - to continue the F1 parallel look at Ricciardo in this year's IRBR-Renault. I think he's WDC material but he was scuttering around the midfield for most of the year.

That's true. It can be an interesting dynamic actually. Good AI in bad car vs bad AI in good car. Who's (or should I say what's) your money on? Or maybe the all rounder team will be the most consistent and win.

You have to be making a looooot of money from races to even consider that. I'm willing to bet it's not going to happen. By the time AI racing is a big thing (if it ever is) then we'll be watching it trackside in VR.

Yeah, it's certainly going to be a looooong time before these fantasy visions become reality.

Realistically, I think the best way is to make these support races for main racing series, with spec hardware for first year, then use the actual main series' cars eventually. You will be able to compare directly the quality of racing between human vs AI.

 

[NW48]

So, how will they throw caps at each other in the pre-podium room?

 

[TenEightyOne]

So, how will they throw caps at each other in the pre-podium room?

They'll be on flappy robot arms, you'll be able to use the app to vote and give your preferred "driver" more cap-slapping strength.