Forgot your password?
typodupeerror
AI Transportation China Google Robotics News

China Catches Up With Google's Driverless Car 258

Posted by timothy
from the OK-but-it's-just-following-the-wall dept.
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 ..."
This discussion has been archived. No new comments can be posted.

China Catches Up With Google's Driverless Car

Comments Filter:
  • by Penguinshit (591885) on Thursday August 11, 2011 @06:37PM (#37063246) Homepage Journal
    Apparently they already have driverless high-speed trains.
  • by Trepidity (597) <delirium-slashdot@@@hackish...org> on Thursday August 11, 2011 @06:44PM (#37063306)

    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.

    • by danlip (737336) on Thursday August 11, 2011 @06:54PM (#37063396)

      Even if the robot car is 100 times less likely to be in an accident, the first time someone gets hurt by one there will be a huge outcry and a lawsuit.

      • by Dishevel (1105119)

        The fact that large numbers of people are self-centered, greedy, and stupid is no reason to stop the progression of technology.

        • The fact that large numbers of people are self-centered, greedy, and stupid is no reason to stop the progression of technology.

          I agree but it does tend to slow things down. Oh well, at least nothing has stopped the progression of technology so far.

      • by geekoid (135745)

        Why? It won't be any different then any other accident. How is this different then say, brake failure?

        • by hedwards (940851)

          I'm guessing that this is going to be like when Toyota was having problems with unexplained acceleration a while back. The matter took on a life of its own due to incompetence on Toyota's part in how they handled the programming.

          They did ultimately get a clean bill of health on that aspect of it, but I don't think the problem was ever really solved. Which is different from break failure which is fundamentally a much easier to investigate problem.

        • by Macrat (638047)

          Why? It won't be any different then any other accident. How is this different then say, brake failure?

          It makes a better headline on Faux News.

        • by danlip (737336)

          The answer to that question is not rational, it has to do with how human beings react to things they don't really understand. Robots and AI definitely fall into that category, and seem particularly scary to many people.

      • by mysidia (191772) *

        , the first time someone gets hurt by one there will be a huge outcry and a lawsuit.

        There will be an outcry, until the video footage recorded in the robot car's black box shows that a human involved caused the accident, not the machine.

        • by Lumpy (12016)

          Actually a lot of people are doing this as well. I bought a video black box for my car. It's a smallish radar detector like device that mounts behind my mirror. when I hit the brakes or if an impact is registered it write protects that file and the next recordings until it fills the device.. I have lots of entertaining video clips of idiot moves I have seen on my commute.

          search amazon.com for accident recorder. better ones are 720P HD and will use MicroSDHC cards to record about 8 hours worth of video b

    • In particular, if we were to take a realistic bar for safety--- beating the average human driver--- the bar is actually pretty low,

      You are approaching an intersection with a stopsign, and arrive at the same time as someone to your left. By the law, you have the right of way. However, the person to your left has started drifting forward.

      Will the computer system catch that? Or what about when someone is attempting to merge and has indicated by glancing your way?

      Its not as easy as youre making it, either, there are a lot of cues on the road that need to be followed so long as other falliable humans are on the road. A good deal of safe

      • by geekoid (135745)

        Those problems have been solved, do try to keep up.

      • Yes the robot car could catch someone moving when they shouldn't be.

        And in Google's implementation, the car knows what a merging lane is and where it is so it could very easily allow in another car as it would be expecting it.

        Now if robot cars become the norm it gets easier still. The computers can just talk to each other to say 'can I merge?'

        • by hedwards (940851)

          This is one of the bigger limitations with human drivers. We can't simultaneously be looking where we're going and looking to see that we can safely change lanes. We have to settle for looking back and forth, which also gives momentary stretches where we aren't looking in either direction.

          A robotic driver could be looking in both directions and as you suggest actually asking for permission. Presumably, that would allow the other drivers the chance to adjust slightly to allow for a safer lane change. Which t

          • by Lumpy (12016)

            In Detroit Directional signals are a sign of weakness... If you use it you will cause someone to FLOOR IT and close the spot you were looking at.

        • Yes the robot car could catch someone moving when they shouldn't be.

          How does it determine whether the other person is looking at you, waiting for you to go, with his foot off of the brake, or actually preparing to move before you?

          Im not saying these are unsolveables, but theyre a heck of a lot more complicated than gp was pretending. There are all sorts of cues that we pick up on that a computer would need to understand.

          • by Jeremi (14640)

            How does it determine whether the other person is looking at you, waiting for you to go, with his foot off of the brake, or actually preparing to move before you?

            I don't think it can -- we're a long way from a computer being able to read a driver's intentions from looking at his (distant) face behind a window.

            On the other hand, I don't think it really has to. As long as its reaction times are quick enough that it can stop itself before the human's car and the automatic car collide, that will probably be sufficient to avoid accidents. (if perhaps not sufficient to avoid getting the finger from the human)

        • by mysidia (191772) *

          Now if robot cars become the norm it gets easier still. The computers can just talk to each other to say 'can I merge?'

          If robot cars become the norm, we can get rid of 'traffic lights', stop signs, etc, and use a network protocol to determine which cars get to enter the intersection in which order, in order to optimize the aggregate cars-per-second rate of the road system.

          Speed limits can also be determined by a safety algorithm, and we won't need traffic cops anymore, which will save government a

          • by Jeremi (14640)

            If robot cars become the norm, we can get rid of 'traffic lights', stop signs, etc, and use a network protocol to determine which cars get to enter the intersection in which order, in order to optimize the aggregate cars-per-second rate of the road system.

            You won't be able to get rid of traffic lights (etc) until all manually-driven cars have been banned from public streets. Which will happen sometime between "when your great grandchildren are old", and "never".

            And even then, there would still be pedestrians, bicycles, motorcycles, etc, to worry about. So I think traffic lights won't be going away anytime soon.

            • by mysidia (191772) *

              And even then, there would still be pedestrians, bicycles, motorcycles, etc, to worry about. So I think traffic lights won't be going away anytime soon.

              Motorcycles will be robotic too. Bicycle riders will be required to be equipped with an electronic interface to the vehicle network. Their bike will identify them to the other vehicles as a rider and tell them when they can stop/go; or they just won't be allowed on the road.

              Lack of traffic lights does not preclude the concept of walk/no walk s

              • by Jeremi (14640)

                As soon as it's shown that all the accidents occuring involve human drivers; government will identify it as a safety feature, and require all new cars be fully robotic within some number of years.

                Dunno what country you're in, but here in the USA I can just imagine how the "Tea Party" types (or their intellectual descendants) would react to that proposal. Good luck! :^D

      • Seems like the problems in the examples you give are based on humans being bad drivers and are the cause of accidents now. The last example you give also depends on the computer acting like a person in the first place. A computer would always assume the car in front could stop at any time and thus will always follow at a safe distance. There's no need to pump the brakes (slow down) because unlike a person, the computer won't be following too close to begin with.
    • Speaking of a social and political problem, AI assisted and/or taxied driving won't become mainstream. Not because of technology, but because of liability. So tell me, when one of these units is involved in the death of a fellow motorist or pedestrian, who's to blame? Who do you think the lawyers are going to go after? The group that has the most money, that's who.

    • 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.

      Per mile driven, just what is the safety record of your average human driver in the USA?

      Based on a quickie check of the statistics, looks like 0.08 accidents per driver per year, on average. Or 0.00025 fatal accidents per driver per year.

      Doesn't look like human drivers are really that bad, when they drive most every day, a

      • by hedwards (940851)

        There are tens of thousands of road related fatalities in the US every year and often times the number is similar to the total number of American servicemen that died during the Vietnam conflict. I'd say there's plenty of room for improvement.

        What's worse is that those are deaths that didn't need to happen, rarely if ever are those deaths that genuinely couldn't be avoided. They tend to be things like drinking, distraction, falling asleep behind the wheel, running a red light etc.

        So, I'd have to say that it

  • 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.
    • Re: (Score:2, Funny)

      by siddesu (698447)
      Yes, it can, except in bad weather. When there is a thunderstorm, the robotic car is prone to rear-end other vehicles and then use its manipulators to bury the evidence.
    • by MaWeiTao (908546)

      I've heard of this sort of thing happening in Taiwan, but specifically with truck drivers. For whatever reason they tend to be a seedy bunch and it's worse in China. My understanding is that the laws pertaining to this sort of thing have been changed. But it all could have been a myth or based on a single incident.

    • by poity (465672)

      Yes, this is because death compensation is capped, whereas a lifetime's hospital fees are not.

      • by poity (465672)

        Sorry, "capped" was imprecise. There's a standard for death compensation -- couple hundred thousand RMB -- and you can either accept it or not.

    • That's bullshit, and a common urban legend told about everyone. The main reason: the courts in China suck. Rich people get away with murder and everything else. There's an exception made if there's a tragedy that requires a scapegoat, such as poisoned milk killing a bunch of suckling infants. But generally, to a degree much more severe than our own, the Chinese court system is skewed to the rich.

    • by FhnuZoag (875558)
      Do you have a source for this? It sounds a lot like an urban legend.
    • by mathfeel (937008)

      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.

      Or he gets out of the car and stabs you to death when he notices you eyeing his license plate: http://www.nytimes.com/2011/06/08/world/asia/08china.html [nytimes.com]

  • by 93 Escort Wagon (326346) on Thursday August 11, 2011 @06:51PM (#37063360)

    If Star Trek (TOS) was made in the current age rather than the 1960s, Pavel Chekov would've been Chinese rather than Russian.

  • 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 obje

    • by wvmarle (1070040)

      Automated expressway driving isn't that hard.

      You obviously have no experience with Chinese-style driving. At all.

    • by Dahamma (304068)

      More than that, automated driving on an expressway (especially encountering only 67 cars in 3 1/2 hours!) is practically supported in cars that are already in production. Lexus has a reliable lane departure alert that could pretty easily be hooked up to the steering system, and a bunch of manufacturers have collision avoidance systems that activate brakes, as well as adaptive cruise control, etc...

  • Given how Asians drive, I'd say this is a great idea.

  • 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 [youtube.com] without incident, it can handle anything the U.S. can throw at it.
    • by PPH (736903)
      I'd say that, given China's broader definition of the act of driving, this is actually a pretty trivial accomplishment.
  • Its China... (Score:3, Insightful)

    by Wyatt Earp (1029) on Thursday August 11, 2011 @07:24PM (#37063662)

    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 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.

    • by Jeremi (14640)

      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 think such a system would depend on all of the cars' software and hardware (speed/location/distance sensors, etc) working accurately at all times.

      All it would take is one robot-jalopy's speedometer or GPS to be off by a few percent to cause a ginormous accident, with fatalities all around.

      Therefore I would imagine a system like this would start out scary for us old timers, and quickly become terrifying for everyone involved. Every trip through an intersection would be a high-speed game of Russian roulett

  • Hopefully the Chinese driverless car fairs better than the Chinese Bullet trains. No one needs another 40 dead.

  • Just wondering, this is China we're talking about (and it's not a racist statement b/c I'm ethnically Chinese.)

  • Will it slam into the car infront of it, and then get burried by the government?

    No thanks. I'll take Google's approach.

  • 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.

Too much of everything is just enough. -- Bob Wier

Working...