Slashdot Mirror
China Catches Up With Google's Driverless Car
mikejuk writes "While Google makes headlines with its driverless car and even manages to lobby Nevada to legalize driverless cars on the public road — China quietly pushes ahead on its own. A driverless car navigated 286km of expressway all on its own. Using nothing but a pair of video cameras and laser rangefinders, i.e. no GPS, it managed to arrive safely even through fog. The computer vision based approach means that at the moment it can only drive during daylight hours. Google might need to speed up ..."
Apparently they already have driverless high-speed trains.
I have something in common with Stephen Hawking...
It's pretty well known that China has been sending spies disguised as academic scholars/PhD students to appropriate information on research projects conducted in the US (professors in my school had been questioned by the FBI and were advised to be careful of these disguised "students"). I wouldn't be surprised if some of the technologies used were stolen from research projects conducted in the US.
They catch up all right, whether by their own effort, I don't know.
Automatically driving a car isn't easy per se, but it's not anywhere near the hardest AI problems we have. In particular, if we were to take a realistic bar for safety--- beating the average human driver--- the bar is actually pretty low, because the average safety record of human drivers is pretty shitty. A robot driver could just not speed and drive relatively defensively, and that alone would give it a big built-in accident-rate advantage, even if its raw skill was worse than a typical human driver.
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
Hello, roboticist here. I'd like to ask you a question: how were power steering, cruise control, anti-lock breaks, fuel injection and collision avoidance radar tested before it was introduced to the commercial car market? When you've answered that question, I'd like to ask you how robotic cars are substantially different in terms of 'experimentation'.
Scientists point out problems, engineers fix them
altslashdot.org: The future of slashdot.
which is why the driverless cars have to date been involved in only one minor fender bender which was covered by the driverless cars insurance and *gasp* this is such a evil experiment for the helpless human drivers of the other cars on the road.
no doubt this evil experiment will one day kill all humans on any road everywhere. it must be stopped. where do i sign up ?
There are a lot less repercussions of a horrible crash in China than in the US. We're slow because we aren't willing to risk lives.
I've heard that in China, sometimes richer people drive cars while poorer people ride bicycles. If a car hits a bike rider, the bike rider can sue for damages. Thus, it can be advantageous, and it's allegedly common, for a car driver to accidentally hit a biker, back up, and run him over again to finish him off. I wonder if and when some company (maybe Google, maybe not) will have cars that do this.
He once inserted random mutations into his code, just so he could have the experience of debugging.
Is this a reference to the research vehicle sharing the road with unsuspecting motorists? If so, I agree that could be a problem.
On the other hand, those fatal "automobile-travel systems" are fatal when humans are behind the wheel. One factor was cited in the article: reaction times. Human motorists also tend to violate traffic laws at whim and make judgement calls that are contradictory to best practices. That isn't a lack of ethics (as in the case of a machine), it is contrary to ethics.
If Star Trek (TOS) was made in the current age rather than the 1960s, Pavel Chekov would've been Chinese rather than Russian.
#DeleteChrome
Slashdot: "China Catches Up With Google's Driverless Car"
TFA: "However Google's car has logged 140,000 miles with only two minor accidents to its name and one of those was caused by a human driver. It will take some effort to match this performance."
I turn up the gain on my Logitech web cam and I can see at night just fine with only the street light for illumination.
Let me know when they get to handling pedestrians, traffic lights, cyclists, areas with different speed limits, yields, and turns in intersections.
its experimental by its very nature. you dont experiment on people without their consent.
They couldn't use miles, they had to extend it by using a unit of measure that makes everyone think it's longer than it really is. We know the truth.
It would be pretty frightening to have a machine steering a car, instead of a good old reliable human who never gets drunk, tired, angry or distracted, and never has a heart attack or a sudden seizure.
Shady Rat at work?
nice to know that the robot car people have, basically, no ethics whatsoever, considering that automobile-travel systems have killed more people than terrorism.
This is China; I haven't been there in a few years, but I think you'd be hard-pressed to build a machine that drives worse than the typical driver I saw on the roads there.
Im pretty sure they didnt just throw them on the highway with other cars prior to testing.
Just because it goes fast, doesn't mean it's caught up to anything. Have you seen the recent news about Chinese bullet trains?
If you don't care about the lives or safety of people, a whole lot more can be done quicker! Right on, China, thanks for showing us the way.
Google's driverless car was just in the news for crashing into a Prius - I was assuming that this headline meant that China had a robot car that had done the same, but I guess this is /. and not Fark.
Besides, Google's car can also look up Sarah Connor when it's planning its route...
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
First of all, killed more people then terrorism? Are you from the US government because I'm pretty certain terrorism is a pretty low bar for deaths, On a national scale america had one big one, and next to nothing for almost all of it's history, on a global scale there's probably 15 different illnesses that can surpass terrorism. Secondly we already are risking our lives due to poorly tested and poorly manufactured systems driving our cars. Humans, we give each new human 30 hours in a classroom, 15 minutes with an instructor, and if they pass that test, good to go. No consent from every driver on the road needed. The high death rate of automobiles is exactly why figuring out something better is a goal. 2 fender benders vs 140,000k miles of drive time is far better then the average human being clocks in when it first learns to drive, and unlike humans as the computer based cars gain experience and intelligence, what they learn is automatically implanted into their next of kin. While humans have to re-learn from scratch every generation.
Automated expressway driving isn't that hard. If you have lane holding and radar cruise/braking control, both of which have been sold in production vehicles, that's almost enough. Quite a number of groups in both the US and Europe have done it. It's mostly a sensor problem.
The remaining hard problems in automated driving involve objects that aren't cars. Children, enemy troops, trash on the road, potholes, bicycles, low-hanging wires - stuff like that. That requires more situational awareness and object recognition, which is hard. None of this comes up much in expressway driving.
power steering: human is in control, power steering augments that control. if it fails, the human can still control the cars direction
cruise control: human is mostly in control. if cruise control fails, the human can still control the cars speed
if the anti-lock brakes fail, you just have normal brakes. the human can still stop.
if the fuel injection fails, the car rolls to a stop, as it would with most other kinds of show-stopping engine failure.
if the collision avoidance radar fails, nobody even notices.
in a robot car the human is not in control, the robot is. if it is going at 60 mph, that is 88 feet per second. if the robot malfunctions and jerks the car into the left lane, then the human can take over... but it takes a good portion of a second for the electrochemical message to get from the brain to the hands. in that time the car has have traveled a good portion of that 88 feet. maybe under the wheels of a tractor trailer, maybe into a bus full of nuns.
is the government watching over these robot experiments, to make sure they are done properly? maybe in the united states, but i can assure you that in China, scant attention has been payed to safety, and any whistleblowers have been put in prison (google Xiao Lianhai).
Given how Asians drive, I'd say this is a great idea.
* Carthago Delenda Est *
Looks like a novel approach to adding a video camera into one of the available adaptive cruise control systems out there.
If an A.I. driven car is capable of navigating Chinese traffic without incident, it can handle anything the U.S. can throw at it.
I once took an excursion to Reddit, and later HN. Unlimited up/down voting sucks when dealing with a hive-mind.
or a bicycle path or a number of other transportation solutions? why pour more money into a system that has proven to be so destructive, not only to safety, but to the environment and human health?
its called trains, subways, bicycle paths, etc etc etc.
all it takes is the desire to put down the kool-aid and stop throwing money down the car-hole.
look mommy, an equivocator!
Heres what psychologists have to do before they do an experiment involving humans
This is exactly why the Google car doesn't ask other drivers about their feelings while stopped at lights. The psychologists said it wasn't ethical.
Other than that, the driverless car has as much right to be on the road as any young learner driver just starting out. And how else would this technology ever get tested under real world conditions, because requiring informed consent from all the other drivers is obviously impractical.
I'm sure the Chinese research team didn't send their robot car out on the public highway without having tested it a lot in the lab and on closed tracks first, and that Google's robot car team didn't, and that the people who developed power steering etc. didn't either. My guess is that none of the DARPA Autonomous Vehicle Challenge competitors did either (or at worst, not many of them :-).
And you don't send a robot car out to drive itself without a human along to override its decisions, any more than a responsible adult would send a young human out to drive unsupervised in a public road for the first time. (Some of us humans learned to drive in "driver's-ed" cars that had an extra set of brakes in the front passenger seat so the instructor could stop the car if he had to, while others learned in cars that didn't have that, so the instructor was limited to yelling a lot and grabbing the steering wheel if needed. And lots of us learned to drive in mostly-empty parking lots before going out on the street.) Presumably the Chinese car had a human backup driver who could override the autopilot if necessary.
It's more fun if you can have the backup driver in the right-hand seat and a large dog or a Terminator mannequin in the left-hand seat, but that's strictly optional.
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
For a test like this I'll bet they found a long straightaway with minimal curves, closed the expressway or used one that's brand new and not opened yet, and set this beast free on it.
Nothing at all like what DARPA challenge or Google do with robot cars
I share the road with unpredictable humans every day. Sharing the road with a predictable computer should be no challenge. Some people worry to much. Computers never worry. I think I prefer the computer over you.
"Windows is like the faint smell of piss in a subway: it's there, and there's nothing you can do about it." - Charlie Br
Can these robotic systems anticipate the thing I can as a human driver?
That baseball rolling down someone's driveway... followed 20 feet behind by a small boy who is focused only on getting his ball back. I can see what's going to happen, using human intelligence, and deal with the situation in a pre-emptive manner. Will the robot do that? Will it be aware that the car behind me is being driven by a distracted driver, and allow *more* than the usual following distance between itself and the car in front, so that if the car in front brakes suddenly, the one behind will have time to stop through whatever distracts the driver? Will it detect the moods and driving styles of other human vehicles from observing their driving behavior, and adjust accordingly?
Will it detect that the pickup truck in front on the highway has a very poorly secured load ready to come off at a bad bump, and speed up to get around it and avoid driving behind the dangerous vehicle?
There are a million situations where I think my human brain is going to be a better thing to have than such a robot controlled car.
Since slashdotters like to criticize: my driving record is zero accidents of any type in 34 years since I got my license. If the robot can match that, in as diverse situations as I have driven in, then I'll consider it acceptable to ride in one. Until that point, I will continue to drive my own vehicle, thank you very much.
it makes sense to be a bit more specific than "China". Was it a government programme, aru university or a company? Saying it's from china almost doesn't add any information by itself, besides that someone besides Google did it
If you honestly think you can compare the two and so easily discount googles system, which really isn't "google's", you are sorely mistaken and obviously have not read a single item about their driving system. The google cars have navigated over 140000 miles without a single accident through day and night, sun and snow and even roads that are not really "roads" by most peoples standards. Its great the accomplishment China has achieved but if anyone needs to speed up its everyone else behind Sebastian Thrun's Teams creation from Stanford University. They all have a lot of catching up to do. The GPS does not help when someone or thing runs out in front of the car and yet that car has no issues traversing even heavy pedestrian traffic or a single deer running in front of it. Its also capable of some pretty amazing high speed race car type driving and car stunts that very few people on the planet can actually do and those that can are typically professional drivers with hundreds of hours of practice.
http://www.youtube.com/watch?v=0FHYpzDinjY
I was thinking about all of these futuristic movies with autonomous cars driving on these California like freeways. In reality if all cars were automated and networked you would only need street level crossings of highways. The cars could weave into the cross traffic at full speed without incident. It might be scary for us old timers but not for long.
I love Jesus, except for his foreign policy.
China's self-driving car will also feature a voice interface. When it detects an American accent, it drives you off a cliff.
Relevant reference videos:
http://www.youtube.com/watch?v=vPxinqO3f8o
http://www.youtube.com/watch?v=D36eisnZ3A4
If they really do manage a robotic driver to deal with that mess it will be a miracle of modern engineering to rival anything in the history of mankind.
=Smidge=
Hopefully the Chinese driverless car fairs better than the Chinese Bullet trains. No one needs another 40 dead.
Which makes the tech actually quite impressive, if it managed to survive the horrible, horrible general traffic behaviour there.
Just wondering, this is China we're talking about (and it's not a racist statement b/c I'm ethnically Chinese.)
And yet look how much time, money and effort we spend trying (ineptly) to protect ourselves from terrorism while practically nobody seems to give two shits about traffic safety.
In 2001 alone, traffic related fatalities outnumbered terrorism related fatalities by about 14 to 1. If we include all the years since it's over 130 to 1.
=Smidge=
Will it slam into the car infront of it, and then get burried by the government?
No thanks. I'll take Google's approach.
Whenever there is something goes beyond your imagination on China, I see sore grape comments. Sure, China copies cause China lags behind Western technology in past 100 years. It has to copy to catch up first. Now the situation is changing. To tell some ambition projects (if you are willing to believe): - China lunar mission project is not national imaging project like US/Russia did. The real goal is to obtain helium-3 from the moon which can enable nuclear fusion plants on Earth. This will be ultimate human energy solution cause it's clean and unlimited resource. - China is building world’s biggest radio telescope in Guizhou Province. It's a five-hundred-meter Aperture Spherical radio Telescope (FAST). To be expected completion in 2016. - China possess the world largest high speed train network across country. They are in service!. I knew what you would say about recent accident. Think about it. Japan has no accident cause it has one line only. US has none. Plane can crash. Don't complain high speed train only. It's relative new. I believe high speed train will compete aeroplane in future. I used to ride many times. It's comfort, spacious and quite. You go to train station, buy ticket and go, no airport hassle and not that slow compare to flight. - China is building its own GPS system, Beidou, rivals US one. Guess what, its receiver can send message back to satellite. A lots more...
Everything comes from nothing.
If an A.I. driven car is capable of navigating Chinese traffic [youtube.com] without incident, it can handle anything the U.S. can throw at it.
Except parking, merging or keeping up to the highway speed limit.
Well, if you are actually interested in the science, the research this car is based on can be easily found, I think, using google. Here, read this article (exhibited at the 2008 IEEE computational intelligence conference hosted in Hong Kong), and if you comprehend it, you can implement their procedure yourself: http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4634099&tag=1
I'll trust that robotic driver over the moronic human drivers any day.
What do you think when I say...
Chinese drywall
Chinese milk
Chinese toothpaste
Chinese public transit
Chinese cars
Chinese iPhone
Chinese toys
Chinese scientific research
In China, the land of the One-Child Policy http://en.wikipedia.org/wiki/One-child_policy, Tiananmem Square, http://en.wikipedia.org/wiki/Tiananmen_Square, the phrase safely could mean something very different that it does in Europe or US. How many Chinese can be driven over before the the driverless car is deemed unsafe? http://en.wikipedia.org/wiki/Human_rights_in_the_People's_Republic_of_China
"They that give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety"-B.Franklin
In China, "safe driving" equates to "don't hit anything and don't get hit by anything."
"There is much pleasure to be gained from useless knowledge." - Bertrand Russell.
The number of people that have died from slipping on KY jelly while trying to do Auto-erotic Asphyxia is higher than TERRORISM deaths in the USA.
It's that low of a bar.... but we need to increase security!!!!!
Do not look at laser with remaining good eye.
What? Did China finally manage to crash their robotic car?
Google's driverless car was just in the news for crashing into a Prius
A human was driving the vehicle.
Last I heard google has not commented on the accident. IIRC the car always has a driver but it is not clear that the driver was actively at the controls. Much as aircraft always have a pilot even when taking off, cruising or landing on autopilot.
As a motorcycle rider. I would prefer a highway FULL of robot vehicles than the hand-full of complete idiots on the highway I see each morning that almost run people off the road. This morning I watched a complete idiot pass on the shoulder at 80+mph with his expedition XL because he did not want to wait for the traffic back up. the number of bend over reflector posts behind him made me grin wide knowing that his precious canyonero is now demolished in the front.
Do not look at laser with remaining good eye.
how safe is it? and how much will get covered up? who will be at fault?
Over there they can have the car hit and kill some pay the family off and sweep it under the rug. In the us and other places all it will take is one death to put a stop to this.
Amen! I can't count the number of times I've had with drivers (whilst on my motorcycle) texting or yapping.
"Never let your sense of morals prevent you from doing what is right" - Salvor Hardin
Fog *and* laser rangefinders?
Speaking as someone who used to use laser rangefinders in inclement weather, I have to call *bullshit* on using any kind of laser in a fog.
Having billions of little floating lenses in the air tends to play havoc with getting any kind of reflectivity beyond 20 feet even from a the most expensive retro-refracting prism you can find. I don't care how much money and technical expertise you've got but you can't fight physics.
Fog turns a laser into the light equivalent of a plant sprayer..
--
BMO
I could see daylight needed maybe for color segmentation? Or really far illumination distance..
I'd think, though, a car's lighting at night that's sufficient for a human would be enough for the machine vision system?
Ask them how that bullet train worked out for them...
After watching this simulation of a computerized intersection, I would say that the situation at the street level would mean the end of airbags in cars, and the introduction of anti-pants-shitting technology for the passengers.
The group's videos page with their approach and statistics is quite good as well.
Wrists killing you? Not in 2 weeks. Learn Dvorak.
In communist China, car drives you!
With the avarage poor-to-appalling quality of embedded software in Chinese products, who in their right mind would trust a Chinese self-driving car?
Unlike the summary states the article says: "It also encountered some problems with fog and indistinct road markings."
What can we say... it's Slashdot, actually understanding a text is not a need to submit a piece on it!
...car drives you!
is paying off. Isn't it?
I think it may be more rational than you're making it out to be:
For instance, the brake system is comparatively simple. And you can test it easily. Press pedal, brake activates (or not).
Not so for the robot driving system. It's hard to set up a full test scenario that'll exercise all possible latent bugs in that system. And the software probably won't be open source.
Not only that, but people have learned from their computers and phones that software hangs.
So how are people supposed to trust the robot? Answer: they don't.
I'm not a lawyer, but I play one on the Internet. Blog
Normally these types of tests actually do have drivers in the driver's seat in case they need to take control. Lots of Google's tests have done as such. Just because there is a driver at the ready doesn't mean the vehicle was driven by a human. Being robotic and driver-less in no way needs to require increased danger to the public.
I imagine Google plans on marketing a delivery system for mail, etc.
Does anyone know what their plans are?
Eg. will it ever be open source?
Help eliminate speeding tickets in your lifetime
Because their automated train system works so well?
Would you give up your control of the motorcycle, in exchange for everyone else doing the same?
In China, "safe driving" equates to "don't hit anything and don't get hit by anything."
I'm pretty sure that's what "safe diving" means everywhere in the world.
The car safely arrived at its destination. In other news dissidence have seem to have been in some sort of hit in run by a car speeding at 286kmh.
Back in AI Class when I was in college we discuses the AI for driving on a highway we noted on a good stretch we could drive a car with some simple AI... Heck you can get fairly good if you just put a brick on the petal.
Then they asked how about having you car to drive in New York City... We came up with an algorithm that works just as well as the average human... Put a brick on the petal...
For some reasons American these past couple of generations have been playing it safe, In my opinion too safe, we are loosing a lot innovation to other countries because new innovative ideas are tossed out because there is some odd chance that something dangerous will happen. Other countries for years earlier had cars that can parallel park on their own, they didn't go to the US Market because of the Law Suits that could happen if a kid or a pet was in the sensors blind spots. (which would probably be a humans blind spot). Yes Human lives are valuable and should give due diligence towards keeping them safe, however it has gone too far in many cases (I know if I loss a loved one due to a safety accident I wouldn't care that it is for the greater good) it is preventing the improvement of technology for humans,
If something is so important that you feel the need to post it on the internet... It probably isn't that important.
Actaully Early power steering was much different then it is today's. I had the opportunity to drive my Dads 1968 Caddi. Its power steering was a full power steering so you had no feeling of the road or resistance when you turn the wheel, It was like driving a video game steering wheel. It was later replaced with the partial power steering we have today because it was safer when people could feel and respond to the road.
If something is so important that you feel the need to post it on the internet... It probably isn't that important.
Aside from the fact that Google was founded by and employs many multi-ethnic, immigrant (1st, 2nd, 3rd generation), and foreign nationals... here we are comparing a company to an entire country.
That bodes well in Google's favor, I think.
I8-D
I usually find that it's the parking that usually makes everything worse. There have been times that the distance I would have had to walk to get my car plus the distance I would have had to walk from where I could have found parking to the place I was going, often was close enough to the total distance to get there that driving was just more trouble than it was worth. Either that, or pay $12 to run into a store for a few minutes. Then there is finding parking again once I get back home.
No one has caught up this. Great what china is doing but watch it and weep.
http://www.youtube.com/watch?v=bp9KBrH8H04
A key factor in reaction times is also whether or not the human driver has encountered a particular driving situation before. Do most people know how to properly recover from a loss of traction and avoid an accident? I would suspect not.
One of the positive aspects I can see with driverless cars is that they would have far more information on hand to make a good decision in a split second. They would know exactly how fast each tire is rotating, be able to judge immediately if any tires are losing or have lost grip, sense not just the car right in front of you but all vehicles and roadways surrounding you, and decide instantly the best way to avoid an accident or result in the least-harmful accident possible.
Essentially, people learn by doing and often by making mistakes. In driving, that sort of "learning" can get people hurt or killed, as it's difficult to train people on those rare situations where an instantaneous, instinctive judgment must be made and applied in order to avert disaster. Over time, driverless cars could actually get smarter as they encounter more and more situations and learn to make the right decisions--and share that knowledge across all such vehicles, a distinct advantage over the status quo.
Personally, I would love to be able to sit back and read a book while my car drives me to work.
Check out my world simulator thingy.
Life being as cheap as it seems to be over there, perhaps it's no wonder they neglected to mention 'the midget Chinaman' peering out from behind the grill... :p