That is quite spectacular, but perhaps not for the reasons the organizers intended. This is pretty much the minimal toy problem for self-driving, with a well-defined course with clear markings, with the only real difference being that the car in question is a full-size racing car instead of a toy car with an Arduino on top.
Of course, the physics of dealing with a real high-speed vehicle is substantially more complex (and more high consequence) compared with that of a toy car moving at a few centimeters a second, but it's not, to coin a phrase, rocket science; the physical principles are well-known, and automotive engineering is a whole industry in itself.
I'd be interested in knowing what actual physical testing was carried out before the race, and how much time the teams had to prepare.
>perhaps not for the reasons the organizers intended.
>This is pretty much the minimal toy problem for self-driving,
Why? The website says this:
>Autonomous racing is not only a new sport, but it is a platform for research and development in the field of autonomous vehicle technology.
By competing in these events, researchers and engineers can test and refine their algorithms and systems in situations otherwise impossible to reproduce, helping to advance autonomy.
What insight did you add or do you have that the organizers missed from their own purpose for organizing this event?
It does looks like the teams didn't have enough time to dial in the software to keep the right amount of traction. The cars were quite frequently spinning out.
If you watched the qualification session, you could even see, some cars suddenly doing sudden twitching maneuvers and stranding themselves.
First off, I want to say it's very fun to do this with Super Formula cars on the F1 Abu Dhabi track. Quite strange to see self-driven cars going around a track with no one in the stands. Seems like a future where we outsource everything to machines, including typical events of living.
Of course I don't know the details of the rules and implementation, but this whole exercise seems to be misguided. Sounds like they have teams hand-coding their algorithms with specific inputs/outputs. That's taking us all the way back to the earliest chess playing programs. We have much better methods today. What's lacking is the ability to train via simulation in a rapid manner. We could even collect data from all the human driven cars that race on the Abu Dhabi track as example inputs and outputs, similar to a database of professional chess games played. Of course we'd also have to note the properties of each car and driving conditions associated with the data, but it's a start. I wonder who designed the format, it doesn't seem to be a researcher who really wants to use this to advance self driving on real roads.
This debacle smells of ill-preparation. Seems that none of the teams actually spent time or effort testing their systems in real life before the event.
For me it shows that you need both computer and domain knowledge to solve something with "the tool commuter". Having both is rare in most fields. Looks like they underestimated the breadth and amount of domain knowledge necessary to be good at racing.
Being good at a computer is not equal to being good with a computer (as tool).
They pioneered robot camel jockeys to replace the original jockeys. Not surprising that they would be into this. Having money, an interest in speed, and a blank desert canvas helps.
At least no small children had to be broken on the road to this endeavor. (You may or may not want to look into the camel race jockey history)
I personally found that quite boring after the first 6 seconds, but of course I don't speak for anyone other than myself so maybe I'll soon be the only person in the world who doesn't find that remotely interesting as a spectator sport?
Part of the reason why micromouse is fascinating is not necessarily the "play-by-play" action, but rather the technological and mechanical skill that goes into running them so successfully - sort of like F1 being more engineering-focused than other forms of racing
From what I understand, F1 is about "emphasizing the importance of driver talent and creating a level playing field", hence the ban of tech like traction control. As an engineer, I'd love to see what the technology could do if unleashed. However, it's probably expensive and I doubt will catch the interest of anyone but uberNerds like me.
Did you consider what will happen once we have this working and we add nitro boost, random wrecking balls, surprise ice segments, collapsable terrain and guns?
The teams started working some time around September 2023, judging by timelines shown on the website https://a2rl.io/.
This is not a lot of time and explains why the performance was so underwhelming (eg spinning out, jerking for no reason). Either that or the compute platform may be under powered.
This is definitely not state of the art performance and if the league continues I would expect much more impressive results next time.
I could see a lot of people talking why they would not watch this. Not just in hn. But I think, that at this point this is not conceived as a spectator sport, but rather an experiment in autonomous driving. Even F1 had to do a lot of outreach and PR, just to stay connected to the viewers. I think this may or may not get popular, but it would certainly attract some racing and autonomous driving companies and enthusiasts.
I've always thought one day we will know autonomous driving systems are working pretty well when a vehicle any normal person can drive can complete a lap of the ring faster with no humans inside than with.
It looks like though the cars could mostly cruise around the track autonomously without issue, they couldn't deal with issues like safely rejoining once spinning out.
Also they seems to be spinning out too frequently, apparently the software isn't able to keep the tires in the right temperature window for the right amount of traction. But that may not be a difficult issue to fix.
I'd still prefer drone pilots for this, so there's a human element as well.
But I've been waiting for proper remote control in this space, so we can finally have races where the cars can be smashed to bits by whatever crazy obstacles we can think of, without any risk to the drivers themselves.
I wonder how many tracks are safe enough from crash perspective. But I would love to see modern series where only limit was something like at least 3 or 4 wheels.
Of course, the physics of dealing with a real high-speed vehicle is substantially more complex (and more high consequence) compared with that of a toy car moving at a few centimeters a second, but it's not, to coin a phrase, rocket science; the physical principles are well-known, and automotive engineering is a whole industry in itself.
I'd be interested in knowing what actual physical testing was carried out before the race, and how much time the teams had to prepare.