Tell HN: I'm creating an autonomous racing series | Hacker News Reader

Tell HN: I'm creating an autonomous racing series

selfracingcars.com

127 pointsjoshu59 comments

I'd be interested in a scaled-down version of this, like autonomous RC without the R or the C.

YouTube delivers: https://www.youtube.com/watch?v=JoHfJ6LEKVo

That's the coolest thing I saw all week! A track like that was the first thing I built all by myself when I was a kid.

I agree that would be cool! I actually worked on the earlier version of that as my master's thesis[1]. I think it would be cool to do this in a less controlled environment where the track isn't necessarily preprogrammed. (Currently there is a camera with IR filter above track and an IR light.)

[1] https://www.youtube.com/watch?v=v-oikffGxOo&nohtml5=False

Interesting! We are doing the same thing but with R/C cars to keep the costs down and make it accessible to more people. R/C cars come in una variety of types and can pretty much run anywhere. Programming and setting up a safe racing environment is much simpler because yoy can have a reusable temporary V2I infrastructure setup for a race. The biggest hurdles right now are the price on some of the sensors required and making a programming environment that is accessible to non-programmers.

If this sounds like something you'd like to give access try feel free to drop me access line. This project is in the pre-alpha stage but there is already research underway with live samples (not simulations).

I like it!

There's a professional AI racing series being planned at the moment. The FIA Roborace [0] will be travelling along with the Formula E championship races in a year or two (their initial schedule is very ambitious). But it's quite a different beast, a 2200 lbs downforce racing car that should be able to do 100 mph+.

But this stuff should be doable by amateurs (with a few bucks in their pockets) or student teams. All the best for OP's effort, I look forward to seeing footage from your races!

And if you act quick, you might be able to beat the FIA RoboRace and become the world's first autonomous full-scale racing series!

[0] http://roborace.com/

I always thought it would be fun if you'd split the field and have the two groups race in opposite directions at the same time than reverse after n laps. Would probably be more interesting as it would create some interesting autonomous driving problems (oh look a super fast object heads on)

Yes! There are a bunch of proposals like this. Obstacles, interfering paths, etc.

Split the track into dry, wet, dirt, sand etc.

Another option is figure 8 racing.

If you don't know about it, School Bus figure 8 racing is an (amazing) thing!

https://www.youtube.com/watch?v=MUuWWqRBpwc

I'm reminded of this delightful story:

https://en.wikipedia.org/wiki/How_I_Lost_the_Second_World_Wa...

Make sure you don't accidentally (or intentionally) set it up so one direction is faster!

You should go ahead and create a company for this, look into insurance, get some basic level of marketing up, etc.

Right now it just looks like a website, with no company (nothing in the footer with even an LLC or anything, and "Joshua Schachter is trying to start a low-end autonomous racing series." needs work as an about section), no news section, no press info section, etc.

The domain name is great, but you'll need much more to really get it going.

For examples:

Founding section: https://en.wikipedia.org/wiki/NASCAR

Race sanctioning section: https://en.wikipedia.org/wiki/Indianapolis_500

History section of: https://en.wikipedia.org/wiki/Formula_One

You could probably pose as an interested historian and get information from each of these organizations about how they are setup and some hoops you might have to go through.

I'm familiar with all of this. It's not necessary to run just a track day. Insurance is dealt with already, etc.

There are, however, almost no vehicles actually set up for this, so the first thing that is needed is a place to test the cars and for people doing this to talk to each other.

Nvidia have been promoting something called Roborace, which is running as something to do with the Formula E series. Is this anything to do with that?

https://en.m.wikipedia.org/wiki/Roborace

Nope. Formula E and Roborace are big money and big spectacle. This is more like club racing. If you are building something and have something to test, bring it.

I was present at the first DARPA Grand Challenge. It really had that indy racing feel to it. Teams up all night long in the paddock hacking away on their code and hardware. So many different sizes and shapes of vehicles all designed for the same goal. Well-funded, industry-backed university teams up against garage hobbyists. It really felt like the future was open to all comers.

Now, we have drone racing, autonomous vehicles that have driven millions of miles, sensors are inexpensive and plentiful, machine learning has advanced massively. I think this could be something cool! (It's possible that the fewer rules, the more innovation it would produce.)

I'm planning to eventually model it off of normal racing - classes will constrain and allow different sets of rules. Fewer rules in some ways, more in others. For example, one class might dictate the car itself, or the sensors to use...

(Not that this effort is intended to do so, but)...

Guessing without wrecks, autonomous racing would likely never become mainstream.

Racing is about pushing the limit until something breaks, then easing juust slightly back from that limit. There will always be crashes, equipment failures, and things blowing up.

Normal racing involves a fair bit of collision, too.

There are at least a few teams out there with working robots.

The racetrack is a much more controlled environment than the street, and there is a great deal more safety set up.

Also, this particular raceway is in an isolated area.[1] There's lots of open space if anything goes badly wrong.

[1] https://goo.gl/maps/1oTTBJNGEjL2

Beyond that it's actually Thunderhill West, which is the smaller track to the west. In that photo, it doesn't even exist except as a bunch of dirt trails.

Edit: I think parent means that the wrecking is what people are watching for?

There's not a lot of spectators in club racing. This isn't for spectators anyway.

Exactly, race clubs are for racing, not wrecking.

Mainstream racing is designed to entertain the fan(antic)s - and they watch the races for the wrecks.

Are you from the US?

Most non-US race series (F1, MotoGP etc) don't have lots of crashes. When they do occur they tend to be annoying for spectators because they occur somewhere else on the track and they interrupt the racing.

Not to mention that despite being "exciting", most people would probably prefer they didn't happen, as when they do occur, they can be fatal. I'm talking mostly about F1 here, as that is how I personally feel. I mean, when a crash occurs, naturally I do want to see it, but it can be an exiting race without crashes...

Good points, though... Some of the best known, and as a result watched races are the ones that had wrecks; my opinion.

Personally, I don't watch races, and do either street racing, long distance, or bad weather driving; which is it say that I don't care for race tracks personally.

If the robots don't run a significant risk of crashing, they'll probably lose to those that do.

There will be wrecks (probably unintentional) but why do you think they'll be required to make AI racing mainstream?

I hear this often from people who don't actually watch racing. Yes, wrecks can be spectacular and dangerous, but overall it usually doesn't make a race more interesting. The wreck is over in seconds, the driver usually/hopefully walks away and then there's minutes worth of clean up effort with yellow flags and safety cars afterwards. American broadcasts typically give you ads, but European race broadcasts don't have ads when the race is going on. So you end up watching a bunch of guys with brooms cleaning carbon fiber bits from the track when you could be watching racing instead.

Then there's the cost aspect. For an amateur series like this, the teams can't afford to crash many cars.

Additionally, when there's humans on board there's always the worry about the driver being alright. After seeing a few fatal crashes unfold in live broadcasts, I don't really want to see any more. It puts things into perspective when you realize that these guys risk their lives to provide you with entertainment.

I don't think many "mainstream" racing fans enjoy seeing crashes.

Not a hardcore racing fan but I have a casual interest in all forms of motorsport and I share your sentiments.

On that note, I wonder how "self aware" the robot drivers will have to become before people start raising similar ethical concerns about their welfare? Really interesting to ponder.

This will be far more exciting to watch than regular racing. Look at Formula 1, they keep changing the rules in order to gimp the cars for safety reasons. Get rid of the drivers and we can let go of the safety limitations.

Get rid of the drivers and we can let go of the safety limitations

Not entirely. There have been a number of horrifying wrecks where spectators have been hurt and even killed. We need to protect against that.

Example:

The 1955 Le Mans disaster occurred during the 24 Hours of Le Mans motor race in Le Mans, France in June 1955, when a crash caused large fragments of debris to fly into the crowd. Eighty-three spectators and driver Pierre Levegh died and 120 more were injured in the most catastrophic accident in motorsport history.

https://en.wikipedia.org/wiki/1955_Le_Mans_disaster

This is really cool. Will hacking self driving cars to allow "racing mode" be a thing in the future? Concept is really innovative and I think it would show just how good self driving cars are vs manual drivers. Hope the big bois (google, apple etc) will jump on if this becomes a thing.

Audi did this about 6 years ago up Pike's Peak:

http://www.wired.com/2010/11/audis-robotic-car-climbs-pikes-...

Audi also has an RS7 set up for self-driving on a track. It can turn quite good lap times but currently the racing line is just preprogramed and so it can't handle other vehicles on the track at the same time, but it's still neat progress.

https://www.youtube.com/watch?v=d2k2LpV6H7E https://www.youtube.com/watch?v=eOYsI1cqUrw

Racing isn't the same as driving fast. I don't think there's been any autonomous racing yet - just time trials.

This is more about the engineering challenge of building and controlling the vehicles.

You should combine this with robot wars. Allow the cars to make contact with each other.

On that note (and I know slightly OT) - I think they're really missing a trick with the new "Robot Wars" by not going fully autonomous. I wonder if any level of autonomy is allowed?

Count Zippy in! He'll bring the food. http://meetzippy.com

Also, the SparkFun Autonomous Vehicle Competition is similar: https://www.youtube.com/watch?v=pfJHgSNx_vY https://avc.sparkfun.com/2014

you are welcome to bring it. drop me an email.

I'll ride along in someone's autonomous race taxi. I think we could get Sonoma on board for some hot laps!

Most racetracks are available for rent.

Most racetracks have crazy, crazy insurance policies and the thought of having autonomous cars on the track would send them into a tizzy.

Can there be a minimum average speed past the first lap? Something to preserve the appearance of a race? :)

I am planning for the first event to just be for testing. Some of these folks have never seen a track.

Slick, thanks for pointing it out. Will be a great area to follow in the future. I think the more who participate in various forms, be it full size or R/C scale, can hopefully advance the knowledge and expertise in the field.

Are there any teams, companies, products, etc that I should invite?

George Hotz

They are already attending.

What size/stage are you looking for? There are a lot of groups in this space that are pretty commercial already.

Looking forward to this, will be really interested to see how it turns out!

This is brilliant. Will be watching closely!

Thanks.

I've been thinking about how to do this for quite a while, and it finally occurred to me I don't have to start with a full-on race.

An announcement is not a Show HN. Additionally, you should use the title of the announcement as the submission title.

I'd argue that it's not exactly in the spirit, but definitely worthy of discussion.

Exactly, which is why I am pointing it out so that the post does not get killed on a technicality.

How about we simply change 'show' to 'tell'. That should make it ok.

Thank you!

I've emailed and asked for a retitle.

Reply on news.ycombinator.com