Extreme Robotic Parking
May 10, 2010 2:50 PM Subscribe
So being a pedestrian in a world of self driving cars will be terrifying and unpredictable.
posted by bswinburn at 2:59 PM on May 10, 2010 [2 favorites]
posted by bswinburn at 2:59 PM on May 10, 2010 [2 favorites]
I so want my next car to be able to autonomously power-slide into parking spots, if only so I can live out the blues brothers every day.
posted by selenized at 2:59 PM on May 10, 2010 [2 favorites]
posted by selenized at 2:59 PM on May 10, 2010 [2 favorites]
a real world example of the power of parallel parking computing
posted by hippybear at 3:00 PM on May 10, 2010
posted by hippybear at 3:00 PM on May 10, 2010
Cool, now I want to see this autonomous car refuel itself, change its oil, check its air pressure, and blare its horn repeatedly at the car in front the split second the light turns green.
posted by Elmore at 3:06 PM on May 10, 2010
posted by Elmore at 3:06 PM on May 10, 2010
Then a second car powerslides and double parks the first powerslider...
posted by jefbla at 3:10 PM on May 10, 2010 [1 favorite]
posted by jefbla at 3:10 PM on May 10, 2010 [1 favorite]
The techniques that make programming robot cars easier are not necessarily the most comfortable solutions for the passengers. Where does the Stanford HCI Group fit in to this project?
posted by Chinese Jet Pilot at 3:14 PM on May 10, 2010
posted by Chinese Jet Pilot at 3:14 PM on May 10, 2010
I hope that this bodes well for computer aided course correction on unintentional slide outs. While this is a fun thing to watch, it would really make me happy to see poor weather accidents reduced significantly through the use of autonomous cars. That and traffic, humans are terrible at mitigating traffic congestion and if systems like this become standard in cars I think it would go a long way in keeping everything moving on freeways.
posted by JackarypQQ at 3:16 PM on May 10, 2010 [1 favorite]
posted by JackarypQQ at 3:16 PM on May 10, 2010 [1 favorite]
"Our system is able to ... execute this maneuver with a final position error of about two feet."
Two feet? That's the difference between a good parking job and a torn off bumper.
posted by crunchland at 3:21 PM on May 10, 2010 [13 favorites]
Two feet? That's the difference between a good parking job and a torn off bumper.
posted by crunchland at 3:21 PM on May 10, 2010 [13 favorites]
A smaller version of powerslide turns for the Robot Wall Racers - see the second video on the skid testing.
posted by warbaby at 3:24 PM on May 10, 2010
posted by warbaby at 3:24 PM on May 10, 2010
Considering it's the first functional implementation a margin of error of two feet is pretty impressive. In 2007's "Urban Challenge" those things could barely drive around a corner.
posted by GuyZero at 3:25 PM on May 10, 2010
posted by GuyZero at 3:25 PM on May 10, 2010
Two feet? That's the difference between a good parking job and a torn off bumper.
In Brooklyn, that's the difference between a parking spot and not a parking spot.
posted by Combustible Edison Lighthouse at 3:27 PM on May 10, 2010 [1 favorite]
In Brooklyn, that's the difference between a parking spot and not a parking spot.
posted by Combustible Edison Lighthouse at 3:27 PM on May 10, 2010 [1 favorite]
Why does it need all those LIDAR gizmos? Is image processing that bad?
posted by geoff. at 3:28 PM on May 10, 2010
posted by geoff. at 3:28 PM on May 10, 2010
It needs a horn that plays "Dixie."
posted by brundlefly at 3:42 PM on May 10, 2010 [2 favorites]
posted by brundlefly at 3:42 PM on May 10, 2010 [2 favorites]
Mod note: A couple comments removed. If you need to talk to us about moderation, hit up the contact form or take it to Metatalk, but do not crap about it in a thread afterward.
posted by cortex (staff) at 3:44 PM on May 10, 2010
posted by cortex (staff) at 3:44 PM on May 10, 2010
It needs all those LIDAR gizmos because LIDAR is cheaper than the equivalent image-processing voodoo to get a reasonably good 3d model of your environment. It's not that image processing is 'that bad', though it is a harder task than the ubiquity of animal visual processing would make you think; it's that LIDAR is a better tool for the job.
posted by Fraxas at 3:48 PM on May 10, 2010 [1 favorite]
posted by Fraxas at 3:48 PM on May 10, 2010 [1 favorite]
Seems like one of those experiments where the statement of aims begins with "wouldn't it be awesome if ..."
posted by aeschenkarnos at 3:49 PM on May 10, 2010
posted by aeschenkarnos at 3:49 PM on May 10, 2010
My Blues Brothers automaton set is nearly complete!
posted by cmoj at 3:50 PM on May 10, 2010 [1 favorite]
posted by cmoj at 3:50 PM on May 10, 2010 [1 favorite]
There's someone sitting in the car, which is either (a) great fun or (b) shows that they are a bit concerned about how quickly it might learn
posted by memebake at 3:57 PM on May 10, 2010
posted by memebake at 3:57 PM on May 10, 2010
As with most things, the Japanese have already perfected this technique.
posted by mathowie at 4:02 PM on May 10, 2010 [6 favorites]
posted by mathowie at 4:02 PM on May 10, 2010 [6 favorites]
I like that it has all the Ghostbusters crap on top.
posted by Artw at 4:04 PM on May 10, 2010 [2 favorites]
posted by Artw at 4:04 PM on May 10, 2010 [2 favorites]
A friend and I (in the legendary Grandada of my Uni days) spent the best part of three hours and two sets of tyres trying to perfect the Blues Brothers/Ace Ventura move on a disused part of a new development one afternoon; by parking facing the other direction on the other side of the street. The combination of the new tarmac and the very dusty (it was still a building site further down) covering of the 6 feet near the kerbs meant that it was quite a challenge. Consistent tarmac would have made it easier, as would a decent set-up on the handbrake, but it was great fun even if we did never manage to get the car a consistent, repeatable distance from the kerb*. We either went too slow to turn the car properly on the tarmac (ending up miles from the kerb and not turned the full 180 degrees), or too fast for the sandy/dusty bit to stop the car after the turn and bent wheels or knocking tyres straight off the rim slamming into the kerb and then having to jack the car up to change the outside rear wheel (again) with it partially on the path next to it.
Very frustrating, although excellent fun.
* The idea was that if we could get it relatively reliable, I'd pull the move in College and park outside the main building coffee shop using the technique. It was part of a dare, but we couldn't get the car to perform it reliably. We put considerable thought and science into it - playing with tyre pressures and various speeds and techniques. I could do it much better in my old car (VW Golf), and we concluded that the Granada was just too heavy and just too shitty to do it properly. We decided there was way too much chance of smashing into something expensive or looking a tool, so we dropped the plan.
People like Russ Swift have this down to a significant level of science, though. Maybe this lot should have got him involved and just logged with a data system what he does as a starting point.
posted by Brockles at 4:15 PM on May 10, 2010 [2 favorites]
Very frustrating, although excellent fun.
* The idea was that if we could get it relatively reliable, I'd pull the move in College and park outside the main building coffee shop using the technique. It was part of a dare, but we couldn't get the car to perform it reliably. We put considerable thought and science into it - playing with tyre pressures and various speeds and techniques. I could do it much better in my old car (VW Golf), and we concluded that the Granada was just too heavy and just too shitty to do it properly. We decided there was way too much chance of smashing into something expensive or looking a tool, so we dropped the plan.
People like Russ Swift have this down to a significant level of science, though. Maybe this lot should have got him involved and just logged with a data system what he does as a starting point.
posted by Brockles at 4:15 PM on May 10, 2010 [2 favorites]
Also, the car isn't power-sliding into the spot. It's just using the fact that cars are inherently unstable in reverse and just allowing that instability to coincide with the parking slot. It's an even easier move than the handbrake Blues/Ventura move (considerably so) and as long as you control the speed accurately and have a fixed rate of turn of the wheel, it's really not that hard to do. The defensive driving techniques use this (called a J-turn) as a means of turning around and getting away from a potential situation quickly. It's really easy to learn.
Sliding into a space forwards is much, much harder to do.
posted by Brockles at 4:19 PM on May 10, 2010
Sliding into a space forwards is much, much harder to do.
posted by Brockles at 4:19 PM on May 10, 2010
I do have to admit, at first I thought that the car was going to be parking in the quite-envied Blues Brothers Style... the actual video isn't nearly that thrilling.
posted by hippybear at 4:28 PM on May 10, 2010
posted by hippybear at 4:28 PM on May 10, 2010
Is the car also narrating the video?
Shut up with the physics and show me some robotic power sliding!
posted by Frank Grimes at 4:50 PM on May 10, 2010 [1 favorite]
Shut up with the physics and show me some robotic power sliding!
posted by Frank Grimes at 4:50 PM on May 10, 2010 [1 favorite]
As with most things, the Japanese have already perfected this technique .
Wasn't there a car commercial like this, with some VW? BMW? sliding into an incredibly narrow space with an inch to spare on either side? It's not for Sprint.
posted by longdaysjourney at 5:31 PM on May 10, 2010
Wasn't there a car commercial like this, with some VW? BMW? sliding into an incredibly narrow space with an inch to spare on either side? It's not for Sprint.
posted by longdaysjourney at 5:31 PM on May 10, 2010
So, how many years until we have cars that completely drive themselves? Because driving is boring, and people get mad at me if I play my DS while driving.
JK, I don't have a DS.
posted by mccarty.tim at 5:35 PM on May 10, 2010
JK, I don't have a DS.
posted by mccarty.tim at 5:35 PM on May 10, 2010
The existence of one car that drives itself will happen in a year or two. Everyone owning a self-driving car is a very long way away.
posted by GuyZero at 5:50 PM on May 10, 2010
posted by GuyZero at 5:50 PM on May 10, 2010
This kid, AKA The Coolest Kid on the Planet, can powerslide/park at age 5.
posted by beagle at 5:57 PM on May 10, 2010 [2 favorites]
posted by beagle at 5:57 PM on May 10, 2010 [2 favorites]
It will be a glorious day when my car can drive me to and from work, so I can take a nap and let it deal with merging into and out of jerkass freeway traffic.
That and when my cat can finally learn how to do my laundry.
posted by darkstar at 6:00 PM on May 10, 2010
That and when my cat can finally learn how to do my laundry.
posted by darkstar at 6:00 PM on May 10, 2010
My uncle the bootlegger could do stuff like that all the time and not spill one drop out of the Mason jar...
posted by Ron Thanagar at 6:09 PM on May 10, 2010 [1 favorite]
posted by Ron Thanagar at 6:09 PM on May 10, 2010 [1 favorite]
As with most things, the Japanese have already perfected this technique.
They're Taiwanese
But the Japanese really have already perfected this technique (from 1:08).
posted by armage at 6:13 PM on May 10, 2010
They're Taiwanese
But the Japanese really have already perfected this technique (from 1:08).
posted by armage at 6:13 PM on May 10, 2010
Found it! (Ok, so it was more than a few inches....)
posted by longdaysjourney at 7:18 PM on May 10, 2010
posted by longdaysjourney at 7:18 PM on May 10, 2010
I prefer this approach. Far fewer lines of code to deal with.
posted by webhund at 7:23 PM on May 10, 2010
posted by webhund at 7:23 PM on May 10, 2010
In the exact reverse of this situation, I can do this in Grand Theft Auto.
posted by thecaddy at 7:25 PM on May 10, 2010 [2 favorites]
posted by thecaddy at 7:25 PM on May 10, 2010 [2 favorites]
There's always the kung fu way.
In the exact reverse of this situation, I can do this in Grand Theft Auto.
Nailing power parking is also one of the fun random side things in Burnout Paradise.
posted by kmz at 9:09 PM on May 10, 2010
In the exact reverse of this situation, I can do this in Grand Theft Auto.
Nailing power parking is also one of the fun random side things in Burnout Paradise.
posted by kmz at 9:09 PM on May 10, 2010
In Brooklyn, that's the difference between a parking spot and not a parking spot.
I still remember moving in to my place in Chicago with a roommate who'd been a Chicagoan for decades. I drove up with my car and was looking for parking when I saw a space that was just a bit too small for my car to fit in, or so I thought. My roomie takes a look and says, "let me handle it." So he gets in and somehow he actually made my car fit. I could swear there were only a few inches on either side. Of course in the process there was a lot of bumper to bumper contact. I asked him, "are you really supposed to do that?" He answered, "they're called bumpers, aren't they?" Couldn't really argue with that.
posted by kmz at 9:14 PM on May 10, 2010
I still remember moving in to my place in Chicago with a roommate who'd been a Chicagoan for decades. I drove up with my car and was looking for parking when I saw a space that was just a bit too small for my car to fit in, or so I thought. My roomie takes a look and says, "let me handle it." So he gets in and somehow he actually made my car fit. I could swear there were only a few inches on either side. Of course in the process there was a lot of bumper to bumper contact. I asked him, "are you really supposed to do that?" He answered, "they're called bumpers, aren't they?" Couldn't really argue with that.
posted by kmz at 9:14 PM on May 10, 2010
Ya'll are like, "I can't wait for this shit to happen so I don't have to drive anymore."
And I'm all like, "Man, I wish my car had a switch that would turn off the antilock brakes so that I could consistently drift my car on tarmac."
The day of the autonomous car looms for me as doom to one of my few consistent daily joys. I fucking love driving.
posted by Netzapper at 11:31 PM on May 10, 2010 [1 favorite]
And I'm all like, "Man, I wish my car had a switch that would turn off the antilock brakes so that I could consistently drift my car on tarmac."
The day of the autonomous car looms for me as doom to one of my few consistent daily joys. I fucking love driving.
posted by Netzapper at 11:31 PM on May 10, 2010 [1 favorite]
[Note: Don't tell me to yank the fuse. I want the ABS 95% of the time.]
posted by Netzapper at 11:32 PM on May 10, 2010 [1 favorite]
posted by Netzapper at 11:32 PM on May 10, 2010 [1 favorite]
mathowie and beagle, don't forget the practice run (d'oh!)
posted by gottabefunky at 12:27 AM on May 11, 2010 [1 favorite]
posted by gottabefunky at 12:27 AM on May 11, 2010 [1 favorite]
Sure, you could wait for the day when your robot car can drive you to work while you take a nap, apply makeup, whatever. Or, you could buy a sports car and drive like a maniac to keep tedium at bay and do it right now.
guess which one I chose
posted by indubitable at 5:13 AM on May 11, 2010
guess which one I chose
posted by indubitable at 5:13 AM on May 11, 2010
Won't we silly humans ever learn??
First they walk, then they run, then they drive; they've been flying, but no, now they powerslide...
Killing us was not enough, the final humiliation of taking that last parking space will make us finally start to fight back.
posted by djrock3k at 9:14 AM on May 11, 2010 [1 favorite]
First they walk, then they run, then they drive; they've been flying, but no, now they powerslide...
Killing us was not enough, the final humiliation of taking that last parking space will make us finally start to fight back.
posted by djrock3k at 9:14 AM on May 11, 2010 [1 favorite]
Considering it's the first functional implementation a margin of error of two feet is pretty impressive. In 2007's "Urban Challenge" those things could barely drive around a corner.
posted by GuyZero at 6:25 PM on May 10 [+] [!]
I realize that they're working on a larger algorithm, and that this is just an example of what it can do, but I'm a little incredulous that they're spending their time teaching these things to, of all things, powerslide when they couldn't navigate the course only a few years ago.
posted by crunchland at 9:49 AM on May 11, 2010
posted by GuyZero at 6:25 PM on May 10 [+] [!]
I realize that they're working on a larger algorithm, and that this is just an example of what it can do, but I'm a little incredulous that they're spending their time teaching these things to, of all things, powerslide when they couldn't navigate the course only a few years ago.
posted by crunchland at 9:49 AM on May 11, 2010
Meh. I used to work with engineers who would rather make buttons flash than build reporting tools into accounting products. Engineers, by themselves, have funny notions of feature priority.
posted by GuyZero at 10:30 AM on May 11, 2010
posted by GuyZero at 10:30 AM on May 11, 2010
I wish they would explain the friction problem more, as they appear to be on a flat, uniform surface and friction coefficient is typically a constant(1) , and one that can be calculated based on velocity.
I guess it's not as bad as a recent Mythbusters episode, where they busted the myth that one car hitting a wall at 100 mph is not the equivalent of 2 cars hitting each other at 50 mph. The reason there's four times as much damage at 100 mph is that the car's kenetic energy is not twice that at 50 MPH, but four times as much.(2) I was waiting for someone on the show to pipe in with a math related explanation, and maybe the audience would learn something, but nope. The disappointment in Jamie's eyes was palpable.
(1) Assuming it's kinetic friction (2 surfaces in relative motion) it's Fk = (coefficient) times (Newtonic force) which could have been measured over several runs and averaged out. Maybe this is what they're referring to when they mention the different models the car uses.
(2) Ke = 1/2 m v^2 . or half the product of mass and velocity squared.
but I'm probably wrong, I'm barely passing the class I learned this in.
posted by hellojed at 1:36 PM on May 11, 2010
I guess it's not as bad as a recent Mythbusters episode, where they busted the myth that one car hitting a wall at 100 mph is not the equivalent of 2 cars hitting each other at 50 mph. The reason there's four times as much damage at 100 mph is that the car's kenetic energy is not twice that at 50 MPH, but four times as much.(2) I was waiting for someone on the show to pipe in with a math related explanation, and maybe the audience would learn something, but nope. The disappointment in Jamie's eyes was palpable.
(1) Assuming it's kinetic friction (2 surfaces in relative motion) it's Fk = (coefficient) times (Newtonic force) which could have been measured over several runs and averaged out. Maybe this is what they're referring to when they mention the different models the car uses.
(2) Ke = 1/2 m v^2 . or half the product of mass and velocity squared.
but I'm probably wrong, I'm barely passing the class I learned this in.
posted by hellojed at 1:36 PM on May 11, 2010
I know this guy and was at the conference where it was presented -- he told me that the folks who maintain the car (a) wouldn't let him try it with anything besides traffic cones, and (b) that they will never ever let him do anything like this again cause of the wear and tear he put on the vehicle.
Hearing that made me kind of sad... at least he got a couple of cool videos for his thesis.
hellojed - the deal is that when you're in a skid, there is a continuous transition between normal driving (tires moving in the direction of wheel rotation), and total slip (tires acting like a puck on ice) which is dang near impossible to model -- you're never 100% in either domain. And unfortunately, even normal sliding friction is pretty hard to model because it rarely acts like the pure mathematical models we learn in physics class.
posted by ubermuffin at 10:42 AM on May 12, 2010
Hearing that made me kind of sad... at least he got a couple of cool videos for his thesis.
hellojed - the deal is that when you're in a skid, there is a continuous transition between normal driving (tires moving in the direction of wheel rotation), and total slip (tires acting like a puck on ice) which is dang near impossible to model -- you're never 100% in either domain. And unfortunately, even normal sliding friction is pretty hard to model because it rarely acts like the pure mathematical models we learn in physics class.
posted by ubermuffin at 10:42 AM on May 12, 2010
the deal is that when you're in a skid, there is a continuous transition between normal driving (tires moving in the direction of wheel rotation), and total slip (tires acting like a puck on ice)
Absolutely. The forces and factors involved in tyre slip are so massively difficult to predict. Everything is interconnected and feeds back. You will also never, ever get to the point of the tyre totally slipping except when it is on ice.
Tyre grip varies according to:
Rubber compound
(softness, basically. This changes with temperature quite dramatically)
Slip angle of the tyre surface itself
(this changes as the effective slip angle and size of contact patch that is able to produce that angle varies according to angle of attack of the road and tyre compound. Also, the release of the tyre grip (the 'bounce' from the slip angle distortion to full grip loss) is very hard to predict and model, especially as the rate this grip is lost depends on how the limit of grip was approached - broadside Scandinavian Flick versus pure g-build up through cornering to grip loss threshold. The 'rate the grip is lost' is an odd term, but it is part of the process that produces a different yaw rate of the car as it slides - ie anything from a gentle and steady movement of the rear of the car out of line with the front, versus spinning like a top down the road).
Temperature of the rubber and road surface
(any tyre produces heat as it works and deforms. This heat may be lost, gained or differently distributed in the tyre carcass by sliding - different temperature characteristics may apply for differing sliding styles; ie full power sliding is a completely different, wheel spinning, grip loss than fully locked wheel sliding without rotation).
That's just a few of them. Tyre distortion is different in full grip versus sliding situations, as the grip reduces, the contact patch reduces, so the heat generated changes and the grip reduces, so the tyre slips more, and the heat generated changes.... etc. etc
Another way of looking at it: The tyres ability to maintain a certain slip angle is directly related to tyre temperature, which is related to compound, which means that if the heat generated by the tyre changes then the compound will change and the grip of the tyre will change, which changes the slip angle the tyre is capable of, which changes how much heat that slip angle generates which....
It's a total mind melt trying to work this stuff out. It's not, by any means, your textbook force and friction equation. Not even in the same zip code. Those equations require a constant friction force, and such a thing doesn't exist in tyre dynamics to any accurate degree outside steady state conditions. Sliding and grip loss scenarios are, by definition, the exact opposite of steady state.
I've over simplified this and lost strict accuracy, but the concepts are true. Everything kind of affects everything else, so as soon as you lose steady state driving, all your 'constants' become interconnected variables. There is a reason that we still need skilled drivers in racing and high performance applications. There is, at present, always a time where computers struggle to model things accurately enough to deal with this kind of thing. We're getting closer, but we have been working with computing models and car dynamics for decades, now. It's not that easy to do it accurately.
posted by Brockles at 11:53 AM on May 12, 2010
Absolutely. The forces and factors involved in tyre slip are so massively difficult to predict. Everything is interconnected and feeds back. You will also never, ever get to the point of the tyre totally slipping except when it is on ice.
Tyre grip varies according to:
Rubber compound
(softness, basically. This changes with temperature quite dramatically)
Slip angle of the tyre surface itself
(this changes as the effective slip angle and size of contact patch that is able to produce that angle varies according to angle of attack of the road and tyre compound. Also, the release of the tyre grip (the 'bounce' from the slip angle distortion to full grip loss) is very hard to predict and model, especially as the rate this grip is lost depends on how the limit of grip was approached - broadside Scandinavian Flick versus pure g-build up through cornering to grip loss threshold. The 'rate the grip is lost' is an odd term, but it is part of the process that produces a different yaw rate of the car as it slides - ie anything from a gentle and steady movement of the rear of the car out of line with the front, versus spinning like a top down the road).
Temperature of the rubber and road surface
(any tyre produces heat as it works and deforms. This heat may be lost, gained or differently distributed in the tyre carcass by sliding - different temperature characteristics may apply for differing sliding styles; ie full power sliding is a completely different, wheel spinning, grip loss than fully locked wheel sliding without rotation).
That's just a few of them. Tyre distortion is different in full grip versus sliding situations, as the grip reduces, the contact patch reduces, so the heat generated changes and the grip reduces, so the tyre slips more, and the heat generated changes.... etc. etc
Another way of looking at it: The tyres ability to maintain a certain slip angle is directly related to tyre temperature, which is related to compound, which means that if the heat generated by the tyre changes then the compound will change and the grip of the tyre will change, which changes the slip angle the tyre is capable of, which changes how much heat that slip angle generates which....
It's a total mind melt trying to work this stuff out. It's not, by any means, your textbook force and friction equation. Not even in the same zip code. Those equations require a constant friction force, and such a thing doesn't exist in tyre dynamics to any accurate degree outside steady state conditions. Sliding and grip loss scenarios are, by definition, the exact opposite of steady state.
I've over simplified this and lost strict accuracy, but the concepts are true. Everything kind of affects everything else, so as soon as you lose steady state driving, all your 'constants' become interconnected variables. There is a reason that we still need skilled drivers in racing and high performance applications. There is, at present, always a time where computers struggle to model things accurately enough to deal with this kind of thing. We're getting closer, but we have been working with computing models and car dynamics for decades, now. It's not that easy to do it accurately.
posted by Brockles at 11:53 AM on May 12, 2010
« Older Get outta there! | Ballad of a South African Football Fan Newer »
This thread has been archived and is closed to new comments
I guess I am a below-average Spectrum reader.
posted by GuyZero at 2:54 PM on May 10, 2010 [2 favorites]