Have we dissected this one yet?

[stonebridge20]

"Autonomous Race Cars To Compete At Indianapolis Motor Speedway* | ZeroHedge" https://www.zerohedge.com/technology...ops+to+zero%29

[DanW]

http://archive.google.com/racing/

[Purple Frog]

It's a very serious event with world wide competitors.

Our own Brandon Dixon (multi-time Champion, Team Citation) is mentoring the effort out of the University of Alabama. They scored a high second in the pre-lims. Now they are teamed up with University of Milan (Italy) for the big show.

My understanding is that the challenges make Formula SAE look like a pinewood derby in comparison. Because of factors involved with racing and speeds, its more challenging than self driving street cars.

[billtebbutt]

..I mean, if that's what folks want their lives to be about. Let's make it so that my computer races the car, so that I can get my thrills doing what else exactly? Programming faster/better than a competitor? Faster keystrokes ("My, look at the conditioning of that guy's thumbs!")?

Yeah, a dinosaur. I know. But i can't help but think that simply driving a car is something that people should do for themselves. Its not that difficult


bt

[Rick Kirchner]

My understanding is that the challenges make Formula SAE look like a pinewood derby in comparison. Because of factors involved with racing and speeds, its more challenging than self driving street cars.

Dissection..... Computationally, it's a bit of a different ball game. You know the course, it's tightly defined so you don't have to figure that out. All the other actors are moving in the same direction, nobody coming in from the sides, no pedestrians, animals, etc. You can predict the physics of the vehicles around you - maybe even access their status (state vector) via telemetry rather than compute it via observation. The surface is relatively constant compared to public roads. The organizers could add interest by making this an unknown, either by the rules (can't map the track before hand) or by introducing mobile barriers to defeat that capability.

I think the big differences are in finding the limit and staying within a percent or so of that. Self driving street cars will be dealing with more variables - but their own performance is not generally one of those. They are going to operate well below the limits and let stuff like ABS take over when they find themselves outside the limits unexpectedly. In the world of self-driving street cars, I think you'll see HP drop to where you get maximum efficiency at about 70 mph because you don't need more than that, ever.

But racing - find that limit & stay close to it, deal with tire degradation, brake degradation, dirty air (may drive aero changes to make those effects more predictable), drafting, surface changes like the track just getting greasy when hot (with electrics, oiling the track will be a thing of the past but if there's a crash, there's still debris to deal with). I think a robo car would be much better than a human over a long stint, but the human might out perform it for short periods of time. Take tires out of the equation and I suspect things get much, much easier.

Sensor degradation might be a challenge.

I'd bet each car will run a model of how much grip they can generate and update the model with some kind of measurement maneuver and every now and then. Some aircraft do this, although air is much more consistent than pavement and race cars operate In the air and ON pavement - much more difficult than just one of those environments.

Then of course, there's strategy. Or is there? If everybody decides to run slightly under the limit in order to finish.....does the guy on the pit wall decide to "go fo it" and raise the limit a tad? Does everybody do that until "the big one" and the guy in 15th is far enough back to avoid the pieces?

A game changer in that it's a different game than the purely human variant. Battlebots are interesting in that the bots are analogs for the humans - don't need to put flesh and blood in there to beat the crap out of each other, and geeks can fight. Question is, would anyone watch more than once if the machines were totally AI? How does an announcer break down what is happening if you can't see or really understand what is going on other than to just scratch the surface?

I suspect that if there's no human element there will be little human interest.

[EYERACE]

Sort of reminds of the bit that I don't remember well at all from MAD Magazine many decades ago. I think it was a satire bit involving how everything was being 'dumbed down'.....so building model cars was, instead of being made more intricate, made easier .....and then the next panel showed a cat assembling a model rocket.

I can hear it now........
......but first I've got to establish the theme > Car Shows (where they're simply on display and people sit around all day, not stuff like SEMA) eventually bored me.....and then one notices the-guy there who talks to everyone all day about all the modifications they've made to their car to make it more powerful or more aero or more this&that and also what they're planning to make to their car in the future to the same end - but they never (because they haven't a clue) actually drive the car near the edge of the envelope.
so......even though others involved will really know what's going on.....
soon there will be engineers talking about their 'new relay' or 'better antenna' or laser-what-not....and the engineer will have no actual competition driving experience
but at least the car will indeed be at that edge.

It's only a 'spec' if you know what I mean... and I've not exactly lost my shirt on it and I'm in the red..... but I will admit owning some stock in LAZR who makes this sort of autonomous technology. I got it! ....maybe I'll talk to the-guy about laser sensors at the next Car Show

[Offcamber1]

Has always been about improving the breed, hasn't it? Isn't this yet another example of motorsports assisting in the development and testing of new technologies?

Spectator interest notwithstanding, it has to be an improvement on the bumper car nonsense of the last couple of F1 races.

[stonebridge20]

i will admit owning some stock in lazr who makes this sort of autonomous technology. I got it! ....maybe i'll talk to the-guy about laser sensors at the next car show

mvis

[mikey]

I want to see humans race computers. Engineering and technology is a looooooong ways off from being able to do what Alonso did this weekend for example. Even just knowing exactly how late to brake isn't easily done with sensors and computers. Or what line is fastest in tricky conditions.
you might like to think race tracks are "well defined" but I wholeheartedly disagree. They are full of nuanced details and ever-changing conditions.

and then there are things like tire management.

any automated racing at this juncture is as gimmicky as WWF. But the automotive engineer in me might watch anyway.

[Rick Kirchner]

Well defined as in comparison to public roads. You know exactly where the track is. It can be completely imaged and geo-registered, to astonishing detail, at little cost. It will bring a new series of problems for circuit owners - once they host these kinds of events, any changes to the facility may have consequences for the participants until they's had time to "see" the changes.

No different than humans though. Danny McKeever used to tell a John Force story. He evidently blew up once by waiting too late to shift. His shift point was a porta potty, and it got moved during the weekend. I had a friend that went off in T9 at Willow Springs once. His turn in reference was the alignment of a power line pole and some other item. SCE changed out the poles for fewer, bigger ones....

Tricky conditions will involve judgement, that comes with training, just like people. You have to feed the beast with a lot of data. I could see developers flooding surfaces and sending the car through. It will likely go off. But if you do that several times, it will probably figure it out.

Training is the achilles heel of a lot of these systems. I once reviewed a targeting system that was fantastic at picking out various types of targets using synthetic data. When they had to train it for real, turns out the data collection required the use of systems that usually signaled immediate hostile intent. That was the end of that. "Sorry about WWIII sir, we were just collecting data".

[Purple Frog]

When engineers developed the drone to be flow on Mars they decided to practice flying it in a simulation room that had the same thin atmosphere. The blades needed to be spinning like 2,400 rpms to get lift in those conditions. There was not an engineer that could fly the thing remotely. The input to mind to muscle reaction time was just too long. They kept crashing. A computer could fly it in the same conditions, and on Mars also..

Just saying.

[Rick Kirchner]

Round trip time delay to Mars is 20 sec, so yeah, by the time the guy in Pasadena figures out what he needs to do the bird has run out of flight time. Round trip to earth geosynchronous orbit is about a half-second. I'm trying to remember the human factors studies done in the 50s and 60s, but if memory serves, once you get to about 100ms (depends of course on what you are controlling) it becomes a matter of chasing your tail. That's total delay, so frame rate in the sensor and the operators displays also come into play.

That's why the big drones are not remotely piloted. Once they get BLOS from the controller and start to work off the satellites, changing the flight path is done via mission updates. Moving the sensor I believe is still done in real time but there are ways around the delay. We had to develop some when the mission computers and data busses were so overloaded in the F/A-18 that even controlling things local to the aircraft had unacceptable delays.

[billtebbutt]

When engineers developed the drone to be flow on Mars they decided to practice flying it in a simulation room that had the same thin atmosphere. The blades needed to be spinning like 2,400 rpms to get lift in those conditions. There was not an engineer that could fly the thing remotely. The input to mind to muscle reaction time was just too long. They kept crashing. A computer could fly it in the same conditions, and on Mars also..

Just saying.

So autonomous Indycar racing on Mars! I'm in !!!!

bt

[mmi16]

So autonomous Indycar racing on Mars! I'm in !!!!

bt

Engineers for a long time have felt the driver was what was holding their genius at bay.

[Peter Olivola]

Engineers for a long time have felt the driver was what was holding their genius at bay.

They've barely surpassed being beaten by your grandmother.

[swiftdrivr]

As a tool to generate interest and train computer geeks / engineers, it makes sense, and those who build the cars will attend and be into it. Their parents will attend, and act "into it." The rest of us will watch for a minute or two, then change channels, kind of like the Olympic shot-put competition. "Oh, they spin around and throw a ball. Okay" [change channel] On the other hand, initially, they will crash a lot, so NASCAR fans might watch for ten minutes.

[racingflyboy]

..

Yeah, a dinosaur. I know. But i can't help but think that simply driving a car is something that people should do for themselves. Its not that difficult


bt

One would think. But, if you've ever observed people trying to drive on the public roads, it appears to me that driving must be one of the most difficult tasks a human can perform!

[EYERACE]

As a tool to generate interest and train computer geeks / engineers, it makes sense, and those who build the cars will attend and be into it. Their parents will attend, and act "into it." The rest of us will watch for a minute or two, then change channels, kind of like the Olympic shot-put competition. "Oh, they spin around and throw a ball. Okay" [change channel] On the other hand, initially, they will crash a lot, so NASCAR fans might watch for ten minutes.

The NAZ comment gave me a great laugh. Thanks.