Slashdot Mirror
VW Raises the Bar for Self-Driving Vehicles
Old Man Kensey writes "According to the UK Daily Mail, VW has produced a prototype Golf (code-named "53 plus 1" in a reference to Herbie the Love Bug) that successfully steers and accelerates itself at speeds up to 150 MPH on tracks designed on the spot without pre-programming. It sounds almost too good to be true given some of the problems CMU's prototype has had over the years, but perhaps VW has learned from and extended CMU's research (and within-an-inch GPS positioning probably helps too)."
"GPS accuracy can be improved further, to about 1 cm (half an inch) over short distances, using techniques such as Differential GPS (DGPS)." - Wikipedia
This is a good illustration of why research funded by a corperation is more likely to achieve results than that of academics. Academics are free to pursue whatever is most interesting as they work, and it is ok to get off on a tangent as long as some papers come out of it. However if you work for a company you need to get results, hence this car. Of course this model doesn't work quite as well for theoretical physics, but well enough for the computer science. I suspect we would have AI already if it could be turned directly into a product.
Philosophy.
Self-Driving Vehicle promptly hits the bar, gets thoroughly oiled and rolls off into the red light district looking for a "service".
"I've got more toys than Teruhisa Kitahara."
And that bug is probably fixed by now, but the problem is, how do we determine we worked out all the bugs? We can't even do that with Linux/Windows/Anything. The closest we come to that in the OS world is a microkernel with only a few thousands lines of code and controlled input.
But how do we ever determine a program that learns and is subject to varying, uncontrolled data inputs is bug free? You can't and I wouldn't want to see the first literal blue screen of death when it happens.
I don't want to sound like a luddite, but the article mentions that planes have been flying autopilot (did they forget to mention landing/taking off is still done by the pilot) since the 1970s. But I believe we'll have flying cars before self-driving* cars because the problem is several hundreds of a magnitude easier in empty 3D space where all you have to do is stay high enough off the ground and avoid collisions via radar/whatnot.
*The only way is I see anything coming close to a self-driving car is on highways where lanes get marked magnetically and driving problem gets reduce to the car having to stay X feet behind the car in front of it.
I have just read about 53+1 the other day (can't rembember where, tough) 53+1 is specialized on slalom courses and can navigate them faster than a human driver. The car first runs the course very slowly scanning it, then it has to pause for half an hour when a special software optimizes steering, braking and acceleration points and afterwards it goes around the course faster than a real driver could. The system is NOT flexible, for example when a human suddenly is on the track on the fast lap it will blissfully ignore the humans existance and accelerate right through the human and create quite a mess. The usage seems to be exactly repeatable driving for car or tyre development. Froh
German news magazine "Der Spiegel" has a pretty high credibility and they carry the same story:
0 0.html
http://www.spiegel.de/auto/aktuell/0,1518,424288,
Unfortunately, the article does not seem to available in English.
You're not familiar with Sunday paper journalism in the UK? Nearly every title will, every week, feature some kind of "exclusive" blue-sky puff-piece about a "new" technology or scientific "breakthrough" which is invariably based on studies or announcements made months ago, or is in fact a highly speculative "what-if?" prediction. If the story contains the sentence, "scientist/engineers predict that in ten years' time...", then you know it's probably not worth reading for ten years.
A tunnel?!?! OHhh NOOOOoooooo...!!
Take the cheese to sickbay, the doctor should see it as soon as possible - B'Elanna Torres, "Learning Curve"
Actually - some car-based GPS systems link up to the car's other sensors (accelerometer, speed, steering, parking radar, etc). While they're not accurate for any long distance, they're perfectly good for a short distance (maybe a few hundred metres) and the software in the system can use this info in the temporary absence of a GPS signal.
Well, in summary (I just read the Spiegel article), the car in question first learns the track based on traffic cones. Actually, the only thing this cars knows are traffic cones. A program then runs on the collected data and calculates the "ideal" path. When the finanlly activate the "racing mode", the car "simply" drives the studied track and that *blindly*. There need not to be any traffic cones, and it will not stop if something unexpected happens (so if a rabbit jumps in it's way, the researches will have rabbit for dinner) It does react a bit on the data from the sensors in the racing mode, but it's more for avoiding small variations in the track like a wet spot.
The car itself is pretty much a standard Golf GTI 2.0 Turbo (200HP) and the only thing they changed was stronger braking. They use the default sensors to make the program learn. Also, in the Spiegel article, there is not any mention of GPS.
Oh, and the research isn't intended to make auto-driving cars for you and me. They want to create a way that cars do exactly the same test runs on test-tracks to check the settings of the car. The results would be more reproductible. If anything, this tech is to put test-drivers out of work ;-)
They also mention that some of the tech was derived from a Touareg that they used in a competition of the US Defense Department in the Nevada desert. However, that one had completely different goals.
I'm sorry that I didn't translate the whole thing, but it was just too long.
Ahhh...the great dumpster continuum. Many a free computer will be found there. -- sowth (748135)
Does this mean we won't be seeing the "Drivers wanted" slogan anymore?
This would imply to me that the position has been filled.
I can claim to be a "rocket scientist", at least I have designed systems for satellite control and tracking, and I work for an aerospace company.
You cannot measure a position to within less than a centimeter using GPS. You can design a ranging system that gives you a measurement with enough numbers to represent that precision, but it doesn't mean that you can trust such numbers.
You cannot use GPS to give you better measurements than the accuracy of the GPS constellation orbit determination, and the satellites' positions vary more or less randomly due to residual atmosphere, solar wind, and solar radiation pressure. The end result is that GPS cannot give any reliable measurement to less than 10 cm, and one meter is closer to the best that one can accomplish in practical situations.
A more accurate system than GPS is LAGEOS, which has satellites that are much heavier and smaller than the GPS satellites. They are basically brass balls covered with mirrors. Because of that higher density, LAGEOS satellites suffer less perturbation from non-gravitational solar and atmospheric effects. However, the equipment for doing ranging with LAGEOS satellites is not portable, it's meant for geodesy studies, not navigation.
A good overview of different satellite ranging systems can be found in "Satellite Orbits", by O. Montenbruck and E. Gill, ISBN 3-540-67280-X, and here is a Wikipedia link for the most accurate satellite ranging systems.
Yes, you can. I just woke up, but I'll see if I can explain.
In the case of DGPS, the reference station uses its surveyed coordinate to difference the time encoded in the signals it is receiving against the time it would expect given an estimation of where the satellite is. So any error in the satellite's predicted position is lumped in with all the other naturally occuring forms of error.
In the case of RTK, or other forms of relative carrier phase positioning, the system attempts to determine and track the difference in the number of cycles of the carrier wave of the GPS signal between the base and the satellite and the rover and the satellite. This number multiplied by the length of the carrier wave, 19cm for L1 signals, gives you the length of one side of a triangle between the base station, the rover, and the point between the rover and satellite that is as far from the satellite as the base station is. So, the exact position of the satellite is not as important as the sight line vector the satellite forms against the base line between the base station and rover. And given the great distance of the satellite from the typical base station and rover, jitter in the satellite's position doesn't change that vector much.
In conclusion, given the advances in relative positioning, limiting factors on GPS positioning today are the accuracy of the survey points, the ability of the electronics to precisely measure the carrier phase/doppler of the GPS signal, the quality of the clock in the GPS unit and the speed/accuracy of the algorithms that determine the carrier cycle count difference.
As far as I know that requires a DPGS like system on the track with extra real time feedback to the car.
So they are cheating if you consider the real world.
I've been in a car that could drive its self on one very well surveyed road. If it got confused it would beep and assume the human was in control within a second. The internal guidance system alone cost over 1/3 of a million dollars and it used several different GPS systems to cross check the fiber gyro.
The only way cars are going to take over for driving the mini-van in place of the drive soccer mom is if there is a serious attempt to clean up the road markings. This means no more optional parking on the side as a road will either be a parking spot or a lane. Signs will need to be redone and cleaned up. The white lines must be far more precise than they are now and more places will need to deal with the yellow centerline (which has now been dropped in the EU even though its the cheapest road safety device ever)
Things have gone a long way. 2 decades ago I had a system that would indicate that a steering adjustment needed to be done. A decade ago there was Miata convertible that could maintain road position and deal with deer. This year we have a VW that can avoid traffic cones. Maybe in a decade we can see a car that can avoid the phone talking, breakfast eating SUV driver.
Well, when I saw that the German -> English translation was almost readable, I decided to run it through English -> Arabic/Chinese/Jananese/Russian/Italian/French -> English.
Stupidity is like nuclear power, it can be used for good or evil. And you don't want to get any on you.
The holy grail is cars that talk to each other to get around more efficiently yet. If the traffic up ahead narrows from four lanes to two because of construction, and car computers can talk to each other and say "Hey, you're two miles back but get ready for this", then orderly traffic flow can be maintained as the cars merge into the remaining lanes and decelerate. This in turn saves gas, etc.
Hell, think how much money you'd save if you car just automatically avoided potholes if it could. Tires, struts, shocks, suspension, all those would last much longer. Look at the figures on how much money it costs drivers annually in a city like Baltimore that's infested with chuckholed roads.
-- Old Man Kensey