volk23 writes "Klaus -- pioneered by car maker Volkswagen AG -- drove in public for the first time around a German test circuit using three laser scanners, a stereo camera, video and satellite navigation systems and radar hooked up to a computer, VW said. "
This page describes a car driven by a Linux computer. It's called the ARGO project, and it's in Italy. It's driven a long distance around Italy in what they called the "MilleMiglia in Automatico Tour".
Don't ever expect this in a real situation.
by
brad.hill
·
· Score: 4
Robot drivers may be fine for controlled situations like factory floors or closed courses, but they'll never be able to approach real drivers for real world situations.
Douglas Hofstadter (AI researcher and author of the wonderful Godel, Escher, Bach) comments on this very problem in his latest book, Le Ton Beau De Marot. He says the problem is as difficult as good machine translation, and that it's extremely unlikely a machine will ever do a good job.
Driving a car is actually one of the most complex things we do. An incredible amount of awareness and thought goes into it. Can a computer:
Realize that it's foggy, or likely to be icy and what patches of the road are likely to be icy and adjust appropriately?
Notice a small child at the side of the road and slow down or and be ready to stop immediately?
Notice a basketball being shot at a hoop above a hedge that obscures a driveway and assume that children might be playing behind the hedge and slow down?
Notice a driver late at night that's weaving slightly, pick him out for a drunk and take appropriate action to avoid him?
See the reflection of the eyes of a deer in the bushes and be prepared for it to jump in the road?
See a driver with lots of dents in his car and give him a bit more space?
Judge from the way the wind just blew that big box into the middle of the road that it's empty and safe to hit if stopping suddenly might cause an accident?
See the canoe on the roof of the car ahead of you isn't very securely tied down and pull back or change lanes in case it falls?
These are just a few of the thousands of situations that come up routinely in driving and which try the intuition of even the smartest human drivers every second we're on the road. As much of an accomplishment as this is, it's mostly empty PR. A computer will not be able to safely navigate a car on public streets for a LONG time, if ever.
Thoughts on computer guided vehicles...
by
cr0sh
·
· Score: 4
The problems with creating a computer guided vehicle are enormous. Just getting the computer to recognize the lines, etc of a road without veering and such, and doing it in realtime - that is a huge problem. Doing the same on a normal road when you have four or five (or more) cars around you, with normal humans (and the woman in the Expedition putting on makeup and chatting on the cell phone - yikes!) is near impossible. Then comes avoiding the small stuff (kids, balls, cats - well, maybe not cats - just kidding)...
I think we shouldn't be concentrating on a complete computer guided car concept, but more on an advanced cruise control for vehicles. Front and rear looking radar (maybe side radar too), combined with a simple vision system (a single B/W low-res cam should do OK here), and smart dots - these things could be done in such a way to make an easy an affordable "smart" cruise control for the freeway commute or interstate drive.
The radar portion would be for speed and avoidance control - this portion is already in limited use in luxury vehicles, as well as some commercial vehicles. The cam system would be forward looking, and would be a system to do edge detection - to help keep the car centered going down a lane. The processing power to do the real-time edge detection and centering should be able to be accomplished with current cpu's, at max maybe a small Beowulf style cluster parallel computer (to process in sections of the image - kinda like the 8x8 processing of jpegs). The final portion would be the smart dots...
What I am calling a smart dot here would be the only infrastructure change needed - and if such a thing hasn't already been created and patented, well, it probably should've been. If not, then what follows I am placing under the GPL - the design, and implementation of anything remotely like it (I doubt this will hold up in court, though). Anyhow...
A smart dot would replace those dots that separate the lanes on the freeway. The new dots would be similar, except that they would have a small circuit in them that put out a very low power 900 Mhz signal, that can only be detected by being near to the dot. The lower edge of the dot would be 3M reflective white, the upper top curved portion would remain clear, to focus sunlight onto a small solar cell to recharge a small NiMH batttery that powers the thing. This "lens" would only be about the size of a quarter or so, maybe even a dime size hole would do it. At any rate, the rest of the dot would be the 3M white, so at night (the battery should be able to last through a night of use - this thing will be very low power) it will show up in the headlamps of the car as normal. This same style dot could be used for the frustum/rectangular style dots, with only a few simple mods.
All the dots emit the low power signal, so low you must be within a foot or less to pick the signal up. Sensors on either side of the car would pick the signal up, so that the car knows it is drifting, and can compensate left or right, and update what it is seeing in the camera vision system to account for the drift. These dots would be the last line of defense (well, not the last line - since this IS a cruise control, the driver is still behind the wheel, and can take over in an emergency). The dots could also emit information, such as a certain exit coming up, or cars need to slow down for road work ahead, or merge left/right (they could be programmable as well - maybe these things use a funky iPic, like that webserver). If they really wanted to be cheap, the dots could encase small rare-earth magnets, and hall effect sensors would be placed on the cars (with appropriate compensation for the earth's magnetic field). However, these kind of dots wouldn't be as flexible, though they would cost only fractions of a penny more than the current dots in use, and probably wouldn't take much to retool for manufacturing them, vs the smart dots.
This kind of system, while not as whiz-bang, would solve a LOT of problems (such as people not being able to merge properly, and vehicle density issues vs speed). It might make traffic move smoother during rush hour, and make it easier to get on and off freeways during these times.
It won't allow you to read a paper or take a nap and catch those few extra minutes of sleep while you go to work - but when it comes to cars, I wouldn't trust any system to allow me to do that!
-- Reason is the Path to God - Anon
The real problems are legal.
by
ucblockhead
·
· Score: 4
The real problem with getting automated drivers on the road is legal, not technical. The way the legal system works (at least in the US), any company that sells these is at a huge risk of being sued out of existence. Even if automated cars are one hundred times safer than the average human driver, the lawyers will still pounce on the company that makes these for a huge amount on the first accident involving one. And given the way juries often work, they'd likely win.
That is the trouble with new technologies. The legal system often demands they not just be better, but that they be perfect.
Will the move to automated driving systems make our journeys out on the highways any better though? Will we end up leaning out of the window and yelling "Stupid MS Driver"? Will we have robots giving us the finger (arm, apendage, tentacle if bioengineering gets going) as they overtake us on the hard shoulder? Will people hack their robot driver to go faster, park on rear-bumper doing 100mph or cut more corners in an effort to avoid traffic lights/ pedestrians/ the police? Will we be attacked with a denial of driving attack by some script kiddie? Or worse, have a UCITA-like end-of-license removal where the originator of the driving OS produces a new version and wants us to upgrade? Stay tuned... I like the look of the bicycle personally:-)
Cheers,
Toby Haynes
-- Anything I post is strictly my own thoughts and doesn't
necessarily have anything to do with the opinions of IBM.
reasons Klaus won't work out on American roadways
by
Lurking+Grue
·
· Score: 5
1) Does not properly extend middle finger at random times 2) Actually comes to complete stop at stop signs 3) Signals lane changes 4) Cannot operate a cell phone while driving 5) Does not blare horn 2 seconds after light turns green 6) Refuses to tailgate car in front of it 7) Does not drive 55 mph in school zone 8) Does not drive 15 mph in fast lane 9) Stops for pedestrians in crosswalk 10) Actually passed driving test
Knight Rider Revisited!!!!
by
BRock97
·
· Score: 5
Wow, and people laughed when Kitt the talking car was driving Michael Knight around while he slept! We have arrived people!
Michael, I have some bad news.
What is it Klaus? Are we running behind?
No, but I did hit two pedestrians and drove through a corn field.
But we aren't running behind.
No.
Good then.
Bryan R.
--
Bryan R. The price of freedom is eternal vigilance, or $12.50 as seen on eBay.....
Slashdot Mirror
This page describes a car driven by a Linux computer. It's called the ARGO project, and it's in Italy. It's driven a long distance around Italy in what they called the "MilleMiglia in Automatico Tour".
If tits were wings it'd be flying around.
I guess VW will have to change their slogan.
Douglas Hofstadter (AI researcher and author of the wonderful Godel, Escher, Bach) comments on this very problem in his latest book, Le Ton Beau De Marot. He says the problem is as difficult as good machine translation, and that it's extremely unlikely a machine will ever do a good job.
Driving a car is actually one of the most complex things we do. An incredible amount of awareness and thought goes into it. Can a computer:
Realize that it's foggy, or likely to be icy and what patches of the road are likely to be icy and adjust appropriately?
Notice a small child at the side of the road and slow down or and be ready to stop immediately?
Notice a basketball being shot at a hoop above a hedge that obscures a driveway and assume that children might be playing behind the hedge and slow down?
Notice a driver late at night that's weaving slightly, pick him out for a drunk and take appropriate action to avoid him?
See the reflection of the eyes of a deer in the bushes and be prepared for it to jump in the road?
See a driver with lots of dents in his car and give him a bit more space?
Judge from the way the wind just blew that big box into the middle of the road that it's empty and safe to hit if stopping suddenly might cause an accident?
See the canoe on the roof of the car ahead of you isn't very securely tied down and pull back or change lanes in case it falls?
These are just a few of the thousands of situations that come up routinely in driving and which try the intuition of even the smartest human drivers every second we're on the road. As much of an accomplishment as this is, it's mostly empty PR. A computer will not be able to safely navigate a car on public streets for a LONG time, if ever.
The problems with creating a computer guided vehicle are enormous. Just getting the computer to recognize the lines, etc of a road without veering and such, and doing it in realtime - that is a huge problem. Doing the same on a normal road when you have four or five (or more) cars around you, with normal humans (and the woman in the Expedition putting on makeup and chatting on the cell phone - yikes!) is near impossible. Then comes avoiding the small stuff (kids, balls, cats - well, maybe not cats - just kidding)...
I think we shouldn't be concentrating on a complete computer guided car concept, but more on an advanced cruise control for vehicles. Front and rear looking radar (maybe side radar too), combined with a simple vision system (a single B/W low-res cam should do OK here), and smart dots - these things could be done in such a way to make an easy an affordable "smart" cruise control for the freeway commute or interstate drive.
The radar portion would be for speed and avoidance control - this portion is already in limited use in luxury vehicles, as well as some commercial vehicles. The cam system would be forward looking, and would be a system to do edge detection - to help keep the car centered going down a lane. The processing power to do the real-time edge detection and centering should be able to be accomplished with current cpu's, at max maybe a small Beowulf style cluster parallel computer (to process in sections of the image - kinda like the 8x8 processing of jpegs). The final portion would be the smart dots...
What I am calling a smart dot here would be the only infrastructure change needed - and if such a thing hasn't already been created and patented, well, it probably should've been. If not, then what follows I am placing under the GPL - the design, and implementation of anything remotely like it (I doubt this will hold up in court, though). Anyhow...
A smart dot would replace those dots that separate the lanes on the freeway. The new dots would be similar, except that they would have a small circuit in them that put out a very low power 900 Mhz signal, that can only be detected by being near to the dot. The lower edge of the dot would be 3M reflective white, the upper top curved portion would remain clear, to focus sunlight onto a small solar cell to recharge a small NiMH batttery that powers the thing. This "lens" would only be about the size of a quarter or so, maybe even a dime size hole would do it. At any rate, the rest of the dot would be the 3M white, so at night (the battery should be able to last through a night of use - this thing will be very low power) it will show up in the headlamps of the car as normal. This same style dot could be used for the frustum/rectangular style dots, with only a few simple mods.
All the dots emit the low power signal, so low you must be within a foot or less to pick the signal up. Sensors on either side of the car would pick the signal up, so that the car knows it is drifting, and can compensate left or right, and update what it is seeing in the camera vision system to account for the drift. These dots would be the last line of defense (well, not the last line - since this IS a cruise control, the driver is still behind the wheel, and can take over in an emergency). The dots could also emit information, such as a certain exit coming up, or cars need to slow down for road work ahead, or merge left/right (they could be programmable as well - maybe these things use a funky iPic, like that webserver). If they really wanted to be cheap, the dots could encase small rare-earth magnets, and hall effect sensors would be placed on the cars (with appropriate compensation for the earth's magnetic field). However, these kind of dots wouldn't be as flexible, though they would cost only fractions of a penny more than the current dots in use, and probably wouldn't take much to retool for manufacturing them, vs the smart dots.
This kind of system, while not as whiz-bang, would solve a LOT of problems (such as people not being able to merge properly, and vehicle density issues vs speed). It might make traffic move smoother during rush hour, and make it easier to get on and off freeways during these times.
It won't allow you to read a paper or take a nap and catch those few extra minutes of sleep while you go to work - but when it comes to cars, I wouldn't trust any system to allow me to do that!
Reason is the Path to God - Anon
That is the trouble with new technologies. The legal system often demands they not just be better, but that they be perfect.
The cake is a pie
Will the move to automated driving systems make our journeys out on the highways any better though? Will we end up leaning out of the window and yelling "Stupid MS Driver"? Will we have robots giving us the finger (arm, apendage, tentacle if bioengineering gets going) as they overtake us on the hard shoulder? Will people hack their robot driver to go faster, park on rear-bumper doing 100mph or cut more corners in an effort to avoid traffic lights/ pedestrians/ the police? Will we be attacked with a denial of driving attack by some script kiddie? Or worse, have a UCITA-like end-of-license removal where the originator of the driving OS produces a new version and wants us to upgrade? Stay tuned ... I like the look of the bicycle personally :-)
Cheers,
Toby Haynes
Anything I post is strictly my own thoughts and doesn't necessarily have anything to do with the opinions of IBM.
1) Does not properly extend middle finger at random times
2) Actually comes to complete stop at stop signs
3) Signals lane changes
4) Cannot operate a cell phone while driving
5) Does not blare horn 2 seconds after light turns green
6) Refuses to tailgate car in front of it
7) Does not drive 55 mph in school zone
8) Does not drive 15 mph in fast lane
9) Stops for pedestrians in crosswalk
10) Actually passed driving test
Wow, and people laughed when Kitt the talking car was driving Michael Knight around while he slept! We have arrived people!
Michael, I have some bad news.
What is it Klaus? Are we running behind?
No, but I did hit two pedestrians and drove through a corn field.
But we aren't running behind.
No.
Good then.
Bryan R.
Bryan R.
The price of freedom is eternal vigilance, or $12.50 as seen on eBay.....