Slashdot Mirror
Driverless Cars Will Compete -- But Only With Each Other -- In Formula E Races
Formula E racing pits single-seat electric cars against each other in high-speed track competition, but the cars -- aside from their powertrain -- are conventional enough, complete with a steering wheel and a human at the wheel. Now, though, the Formula E series will also incorporate self-driving cars. From the article: Ten teams, each with two cars, will square off against each other in hour-long races on the same circuits that the Formula E cars will hurtle around. The cars will be the same as the next in order to get the teams’ developers to focus on creating better algorithms and artificial intelligence to win. It takes inspiration from how the Formula E teams were required to run the same cars in the event’s debut season, which meant there was more focus on the development of battery technology.
No driver will save some weight. Or you can lower the position of more battery so that you get some more charge and better weight distribution. And obviously, without a driver, they can pull more G's in turns, nevertheless still limited by tires and downward force.
One concern I have is that without a driver, safety will become less of a concern. Obviously, they also don't want to wreck cars that cost millions of dollars, but a wreck would not incur the loss of a human life. (I hope that's important to racing teams.) However, if these cars will be racing at the same time on the same track as human drivers (even if they're in two separate rankings), then the risk to other drivers will be increased.
The winning team will use Rust to program their car.
It is a systems programming language so it can be used for systems like artificial intelligence driving a car.
It is blazingly fast which is what you need when driving fast.
It prevents segfaults which is good because you don't want a segfault when you are driving!
It guarantees thread safety which means the fabric on the seats of the car will remain clean and intact.
It has zero-cost abstractions which will keep the price of the car down.
It has move semantics which is just what you want from a car: movement!
It has guaranteed memory safety which is good because you don't want your car to forget where it is.
It has threads without data races which is critical because it means that there can be no data racism because there are no races.
It has trait-based generics which are good because they mean that, generically speaking, the car has good traits.
It has pattern matching which is very important because the car needs to differentiate between what is road and what is not road.
It has type inference which is good to have because it can infer what is road and what is not road.
It has minimal runtime which is totally what you want in a race: you want to run it in the shortest time possible!
It has efficient C bindings which means it's as fast as C because most Rust code actually gets written in C and then glued together using Rust because C is a much more useful, efficient, and faster programming language than Rust is.
We are going to remove people from driving cars but insist to put humans into tin cans in the upper atmosphere.
Um, it should be the reverse.
The cars will be programmed specifically for each race, following planned strategies, on painstakingly staked out tracks which will look nothing like a normal road.
Well, other than vehicle control at the limits of traction, machine learning to find out what those limits are, emergency control when limits are exceeded, changing track conditions like rain and obstacle avoidance for when the other cars run into a wall and leave car bits all over your pre-programmed racing line.
But will the self-driving cars be 3D-printed on Mars by Elon Musk?
You are welcome on my lawn.
This is actually a lot more interesting than might first appear. When doing timed laps such as during qualifying, sure, it is pretty much about just optimising a whole bunch of parameters. I would imagine the superior ability of the driverless car to sense grip and slip on each wheel and create a dynamic map of grip around the track would mean it could quite easily beat a human driver without a lot of 'artificial intelligence'.
However, when wheel to wheel racing it is a whole different story. Unless you have artificial overtaking aids (DRS, boost modes) then overtaking a car that has the same performance as you is essentially a mind game. A common strategy is simply to hound the other driver using the draft to get right on their tail, fill their mirrors, and try to pressure them into making a mistake. This used to be quite common in F1 before we had fuel and tyre restrictions and DRS. It was a real test of a driver's mental strength to have to perform under that sort of direct pressure.
Another way is to play games with the other driver. For example, on a corner that allows loose racing lines, you can try to trick a car you want to overtake into defending on your weaker line, but then once you've lulled them into believing this is the way they always need to defend, you swap it up. Sometimes this may only get you in position to complete the overtake on a subsequent corner, so these types of moves can be very complex.
However, perhaps the most psychological part of racing, is simply the 'how crazy is the other driver' effect. Sometimes wheel to wheel racing is all about playing a game of chicken into a corner. The most common way is you dive in and rely on the driver you are trying to overtake giving you space, even if it is border line as to whether racing rules would require them to do so. You basically say, I want the corner, you either hit me and we both go out, or you let me through. This plays out in really interesting ways, for example in F1 Bottas and Raikonen have been in this mind battle for the last few races. They had a run in earlier in the year where Raikkonen made a move on Bottas and they collided. Bottas was out of the race and Raikkonen essentially got his place. In a subsequent race, Raikkonen tried a similar move and Bottas didn't give him any space, causing them to both get taken out of the race. Bottas actually said after the race that if he'd given Raikkonen space then everyone would just try that pass on him, knowing he is a soft touch.
When we actually have good wheel-to-wheel racing (which sadly F1 has not had for a while now) motorsport does relies heavily on human performance and it's faults. That is still why a lot of fans watch it, and I think this AI car thing really misses that point.
Hey, remember that time I posted that Bill Nye should praise Formula E instead of bashing NASCAR and then people just started bitching about how stupid Formula E was?
. . . Suck it!
Sdelat' Ameriku velikoy Snova!
Ok, not at first, because it's a war of the minds of engineers. = Tension!
But that is not what the majority of people want to see. We could test that for example if someone fires up his/her FEA Application and introduces people to the painstaking art of stress&weight optimization of a critical design.
I would watch that for hours, but I bet any of my neighbours would.
Also with control optimization for such driverless cars - to operate near the tipping point when rolling friction tipps to sliding friction.
That would simply be their secret and in consequence kept secret.
Simple Example:
When rolling friction tipps to sliding friction = you hit the brake of a pre ABS-car and your wheels are stopped from rolling and then they only slide.
So where is the fun. The fun in a car race is the human factor or the human failiure!
well, it was really unexpected but they actually managed to invent a race where i'm not secretly hoping the cars fly off the track... ok, well maybe just a little.
Anons need not reply. Questions end with a question mark.
The winner will be the car with the most realistic image of a child about to run into the street painted on the back.
You are not alone. This is not normal. None of this is normal.
Nah. Computational speed isn't really the issue.
Some F1 teams have two-seaters and for a hefty price or if you're a daring VIP you can get a couple of laps around the track driven by an F1 driver (usually test driver since the stars' time is too valuable). Maybe these cars could be used to sell such experiences in an even better format. You could get better visibility than from behind the driver and maybe also have some adjustment for how many percent of maximum speed you want the car to drive you. At "full" speed such an experience makes some people pass out due to the G-forces in some curves.
Duh! It'll be Swift, 'cause they want it to go faster!
(Still on phone, not logged in.)
The task of racing is very different then the task of driving. It requires aggression, not caution. Current driverless cars are all about not crashing, and that is a very different problem.
It's really different. For example, there needs to be a yellow flag mode where cars hold their relative place. When there is an accident with debris on the track, how do you clear the track? Do all the cars stop?
Can the navigation algorithms execute responsively enough at track speeds? Many of the current automomous vehicles use 3D scanners. Can those keep up at high speeds? How does the common vehicle encompass all the different sensors that various competitors want to use?
Look at the results of the DARPA Grand Challenges. The cross desert race was won by the team that manually added a large number of way points, in some sense defeating the intent of the challenge. The more recent robot challenge showed that humanoid robots are not ready for very much besides walking and cannot traverse uneven terrain. Why should race cars be any different?
Still, I think it would be an entertaining first lap where everything goes to hell and the track is strewn with really expensive gear. You could make book on how many cars make it past the first turn. Zero would be a good bet.
Why is Snark Required?
The only winning move is not to play.