I’ve only been to the race once. That said, it seems that Team Nerd could take the Checkered Flag.
Yes.
Are you asking about the capabilities of the vehicles, or the rules of the race?
Absolutely. With a totally predictable roadway, better reaction times, probably better look-ahead awareness, and no driver fatique, seems ideal. Unless another driver ambushed it from behind.
I’m wondering if it could navigate pit road in heavy traffic without losing too much time.
At a plate track like Daytona, the AI would have to be extensively programmed with regards to the draft, which is more of an art than a science, at least as far as the driver is concerned.
Yes, for all the reasons stated: it’s a completely controlled environment with predictable conditions, ideally suited to the fast judgement of a computer.
(Which is why SDC’s have succeeded so far… in flat, stable, modern roadways with a high degree of predictability and few complications. I await the one that can work outside of a San Jose industrial park or 10yo upscale suburb.)
And it should be allowed to use both a turbine engine and ground-sucker secondary motors. 
Drafting is an art for slow, distractable and fatigue-able meat computers. I hesitate to say it would be trivial for a computer-controlled vehicle, but it would be pretty close.
From a couple of stories I heard on the radio last week, I got the idea that the programming in self-driving cars makes them tend to be somewhat defensive and deferential to other drivers. While more aggressive/assertive programming is no doubt available, in a race between a single self-driving car and a bunch of human drivers, I would hesitate to put my money on the robot.
Maybe not the first few iterations. But if allowed to persist very long, the sub-millisecond detection and response times would end up easily trouncing any human driver. You could even program a whole host of the same crisis decisions drivers have to make in contact and best-bad-choice situations.
Street SDCs are being held to the highest, purest form of the Three Laws. A NASCAR robot could work to vastly blunted forms. It would also never tire, never misjudge (to the limits of its sensory input, anyway) and never fail to keep track of the endless details of professional racing.
True. But it’s no harder to program a self-driving car to drive like a ‘roided out redneck in a jacked up pickup with a big Confederate flag a’flyin’.
It’d be more than happy to cut you off with milli-inches to spare as opposed to the actual whole inches the human redneck will leave you.
On paper, sure it could. But in actuality could it? No way. There is only one way to prove it could, and when I see it happen, then I’ll believe it.
No way, no how.
I was thinking specifically of Dale Earnhardt Sr. where his intuition and experience would seem to tell him where the air currents were optimal, or such.
Someday, but not today. Audi has an autonomous car in Northern California that just recently beat a driver - a good amateur but not a professional, I believe - by half a second. But these were timed laps, not two cars driving at the same time.
There’s little question that autonomous driving will get better and better, but we are not to the point of a car racing against dozens of competitors on the same track, let alone win.
Yeah, we are not there, yet. The AI racers would need to be programmed to open a bottle of champagne.
10w-40
Sparkling 10w-40!
The programming includes a few adjustable parameters (where “a few” probably means “hundreds”) that the programmers can tweak from one test drive to another, to see how it changes things. There’s a parameter (probably many of them) to specify how aggressively the car drives. Google sets their cars to higher-aggressive when they do test runs on their tracks, but to low-aggressive when they drive around town.
One example I read about: They noticed that when human-driven cars are stopped at a four-way stop sign, the drivers will start to creep out into the intersection before it’s really their turn. This is one of the ways the drivers negotiate with each other who gets to go next. If a driver doesn’t take the initiative to creep out into the intersection, other driver(s) will and those are the ones that get to go. The timid driver could sit there all day.
So Google has an adjustable parameter in their programming to set their cars to creep out into intersections at a stop sign like that. This is the sort of thing that they will disable (for now anyway) when driving in public, but they enable it for testing on their track.
I have noticed in a couple of these ‘AI driven cars on race-track’ videos I’ve watched that the actually physical control of the cars steering is quite slow, ie: the wheel is turned left or right with a piston-driven arm (I’m not an engineer, I have no idea what the actual term is), which is fine for a smooth and clear track where the AI has time to set the car up for the upcoming corner but might not be feasible for an actual race where it has to react extremely quickly for unpredictable events.
Basically what I’m saying is that the software can analysis and react faster than a human but the physical means of turning that into real movement seemed quite slow. At least to me.
Another test perhaps is to have humans compete against AI driven car’s in computer simulations before trying it out for real? I do know that in most games the AI’s are nixed to a large extent because otherwise the player could never beat them.
There are actuators like that that can move a couple feet in a small fraction of a second with sub-millimeter precision. While pushing a few hundred pounds. IOW, much stronger, faster, and more precise than any “intelligent” hunk-o-meat’s actuators.
Bigger versions of those are used to support aircraft simulators.
I agree that a prototype self-driving car may well have an underpowered actuator as a failsafe so the human driver can always overpower it even if all the automatic cutouts fail.
But, you may also be mistaking smooth control for slow control. Done right, it’ll drive with a lot less steering wheel hunting than you do.