Driverless Roborace car crashes at speed in Buenos Aires | Hacker News Reader

Driverless Roborace car crashes at speed in Buenos Aires

bbc.com

65 pointsZeljkoS26 comments

If there's no crashing, what would be the point of watching?

Seriously, nothing would be more boring than watching 10 cars all driving perfectly in a line. The opportunity for aggressiveness, risk-taking and crashes without human lives ending is what makes this exciting.

I think that is where the different Autonomous systems come into play. Each one may take a slightly different approach to each decision point.

It would be like the difference between Rinaldo Capello and Allan McNish in endurance racing. Both very fast drivers... McNish is much more aggressive, who took more risks and got away with them more often than not verses Capello who was much more conservative.

This is exactly the sort of thing we need to test the limits of robotic vehicles. Nothing spurs innovation like intense competition.

Also think it would be cool if the cars could get points for touching other cars (without crashing, of course) or faking them out with optical distractions or anything else.

Or unleash more obstacles that cars have to avoid or lose points. Fake babies crawling on to the track and such.

I Snow Crash by Neal Stephenson, the mafia run pizza company used some that might be useful in such a race. The pizza delivery cars where specifically designed with a sensor suite (lidar, radar, optical camers, etc..) and some software to find two black dots about 3 or 4 inches apart that moved together.

Once found it determined if those pupils belonged to a driver of a vehicle, then determined if that vehicle was in the way or not. If those pupils did belong to a driver with a vehicle in the way of the pizza delivery, a quick blast of laser light would insure the vehicle couldn't react and the pizza delivery vehicle could pass on whichever side the vehicle didn't swerve.

Add actual weapons fired from car to car and you've got something that I'd pay money to watch.

So pretty much the movie Death Race but with self driving cars.

Might as well go full Full Auto 2 and let people duke it out with them.

I'd pay money to watch that.

I've been tracking this Roborace thing and they're spending a lot of money on bullshit. It's all sizzle and no steak. They have some very slickly produced videos about the passion and drama of racing robots and they're all hilariously anti-climactic.

Methinks the talent shortage in the autonomous driving space is severe enough that anyone capable of developing an (exciting) autonomous race car has more important things to do.

Depends on where their funding is coming from, and what the goals are. If they're being funded by big corporations as both research and PR, then it's not just junk. They're doing good work.

Is doing this on a track too ideal? Would be cool if this was done in super extreme conditions. That R&D would definitely help commercial autonomous cars.

I searched Youtube but could not find the shortest clip of the event.

Are there any?

BBC says they will publish it on Friday on YouTube.

Ya I was hoping for a video on the article itself. No footage of the actual crash =(

I was terrified when I saw the picture of the dog before I read the caption.

The caption is downright hilarious because of just how dry it is.

"The winning Devbot 1 managed to avoid running over a dog"

I thought so too.

"Devbot 2, on the other hand..."

I'd like to know more specifics on how the cars communicate with each other – is the event more a choreographed show than an actual race?

I follow motor sports to an extent, but don't feel too enthused about watching computers race each other (no matter if on track, or virtually), but it would further diminish the “racing” part of it if the machines actually told each other what they were up to next. A more humane way of driving might be more spectacular.

Hopefully it's on te level of what drivers already do. If you see someone is taking you on the inside, you don't close them off in a way to deliberately cause a crash. Basic politeness.

If they communicate so it's all a choreographed dance, then that's stupid and the sport is nigh unwatchable.

There's a reason "best crashes" videos are a whole genre on Youtube. The danger makes motosport interesting.

> you don't close them off in a way to deliberately cause a crash. Basic politeness.

Typically the rules of racing will say you don't block faster cars or deliberately drive dangerously, etc. e.g. Forumla 1 says:

https://www.formula1.com/en/championship/inside-f1/rules-reg...

But yes, this Roborace thing seems pointless unless the cars are made by different teams and have different software, etc.

Apparently that's how it's going to work: https://arstechnica.com/cars/2016/08/meet-devbot-a-self-driv...

My guess is that the "communication" is fairly basic (e.g. position, speed, acceleration - nothing a human driver couldn't observe), rather than information about strategies (e.g. "I am going to attempt to overtake on the inside at the next turn").

Done right, I could see this event being quite spectacular. Vehicles could take bigger risks, and reap bigger rewards (and bigger crashes).

One of the things you could do with Chessmaster was have it play itself and watch, this turned out to be pretty boring. I suspect that having robot cars race would be similarly boring for everyone except the engineers who were enjoying the technical challenges being overcome.

Perhaps it would be boring if all cars are running the same basic strategy. If competing teams get to tune their strategies and counter-strategies I can see this being quite popular. After all, there might be bigger crashes.

This makes me wonder about the possibilities of AI-based racing in the future. Can we make a truly competitive sport of it? Grand Prix style auto racing is a really complex thing, with the most sophisticated cars in the world, and a lot of judgment calls by the team as well as the drivers. Speeds are limited by human reflexes and safety. Theoretically, we could get even faster, more difficult racing with AI. But will AI flatten the curve to where it becomes uninteresting?

AIs are still limited by the capabilities of the hardware, and there are a huge number of variables to deal with. If the teams are properly incentivized to win even if it means occasionally having a crash because the car was operating on the bleeding edge, I think it should stay interesting.

I think the bigger question is whether it's actually fun to watch cars racing without humans in them, no matter how exciting the racing is otherwise.

"DevBot in Driverless mode — scanning the Donington Park circuit in England"

https://www.youtube.com/watch?v=xrfCQ7qwPh8

>at speed

at _low_ speed, no idea why they run at what looks like 1/2 power of human counterparts

Reply on news.ycombinator.com