Slashdot Mirror

← Back to Stories (view on slashdot.org)

Helicopter Crashes While Filming Autonomous Audi

Posted by timothy on from the send-robot-ambulances dept.

telomerewhythere writes "A helicopter commissioned by Audi to film its autonomous Audi TT climbing Pikes Peak crashed early this morning. Four people on board were hurt, the pilot seriously. It's a surreal story — a manned vehicle crashes while the one climbing a mountain driven only by computers and sensors carries on. Here's more on the autonomous Audi, a project undertaken with the help of Stanford University."

12 of 218 comments (clear)

  1. GPS? by Esospopenon · · Score: 2, Interesting

    As mentioned in TFA, they hope to create "autonomous driving systems that will one day be integrated into all vehicles as a safety measure". That being the case, I think they still have a long way to go since they have fitted a $100.000 GPS system for guidance. They also have a driver running the course first so the system can "incorporate human reactions", which probably means "learn when to breake" The real test for this system will be when it can cope with unpredictable situations, like traffic lights and old men with hats.

    1. Re:GPS? by mindriot · · Score: 2, Interesting

      Wow, no you can't. For one thing, you don't have to deal with other cars when you're making an exhibition run up Pikes.

      That depends on what you mean by "taking over". If someone falls asleep or has a heart attack while driving, "taking over" can just mean bringing the vehicle to a controlled stop in a safe location and turning on the hazards. In that particular situation, there also won't be much room for suing anyone if something goes wrong -- because had the vehicle not done anything, the situation would've ended gravely anyway.

      Also, while you may not have to deal with oncoming traffic running up Pikes, you have to have a damn robust and fast perception system that is able to react to its environment quickly and safely ("oh, pothole on the right, better avoid that"), and you need to have a software capable auf autonomously controlling a vehicle in the most extreme situations (such as going round a curve on a dirt road at 60 miles with a hundred-foot drop on one side. I dare say that if you manage that, you'll be doing pretty well in "normal" traffic as well. Combine this with the expertise gained from the DARPA competitions, and that "long way" is already getting shorter.

    2. Re:GPS? by turbidostato · · Score: 2, Interesting

      "Now, if a robot drives a car, he has two options: Follow the law, cause a traffic jam behind himself"

      Provided all other cars were robotical, how could this happen? A traffic jam creates when vehicles reaching a point are faster than vehicles at that point. Provided our car is respecting security distance from a car that was at top legal speed, it would go at top legal speed as well as those after it. No traffic jam is possible. And with regards to speed change and wave effect since robots would have faster and finer-grained reactions, those would be minimized too.

      "(or even provoke somebody to cut into the seemingly extremely long clearance)"

      It can only seem to be to long if you think you can do it better. On one hand, other robotic cars neither would nor could find it "too long" but just "as it must be"; on the other hand, provided robots have better reaction times than humans (and that's a perfectly defensible position) clearance distances could be reduced basically to the dynamic envelope of running vehicles, no need to take human reactions into account.

    3. Re:GPS? by khallow · · Score: 4, Interesting

      Human beings have no business driving. I know this statement bothers a lot of people, but the statistics bear it out. I, for one, will gladly hand over my keys the day I can buy an autonomous vehicle, and never think twice about it. Driving is a chore 99% of the time, and one that I'd be just as happy to turn over to a computerized device as any other chore.

      Human beings have no business being alive either. I think statistics will bear that one out too. Look, I recognize that there are plenty of activities, even plenty of transportation activities that don't require me to be in control. For example, we routinely travel by means that have someone or something else doing the driving (passenger trains, airlines, etc). And these means of travel are usually (at least in the developed world) safer than if I were driving myself. But driving is "do it yourself" mobility. In exchange for a somewhat elevated chance of injury and death, you gain a great deal of freedom.

      Second, driving engages me. It is often fun to drive a car.

      Ultimately, safety is not the key point of driving or for that matter, it isn't always a chore. Else we never would have left the house in the first place. And an autonomous system won't be able to cover all those needs that formerly used a human driver.

    4. Re:GPS? by SparkleMotion88 · · Score: 2, Interesting

      Right, this was probably an expensive INS (e.g. ring laser gyro) + GPS system. I imagine they would use something like this for a prototype, because they can get precise acceleration information without having to recalibrate after changes to the vehicle. The good news is that it is easy to get acceleration data for cars because they only have one axis of rotation, and they have wheels that are always on the ground. For production cars with this sort of technology, they can probably just use the speedometer along with something that measures how much the front wheels are turned. Maybe they will throw in a cheap INS to detect loss of wheel traction or otherwise improve the quality of the speed/acceleration data.

      Of course, this only helps for navigation, which is probably the easiest part of the problem. Collision avoidance would probably be done with cameras, radar, ultrasonic ranging, etc.

  2. Re:Uber-silly by timeOday · · Score: 4, Interesting

    I don't think your ambulatory computers will ever be clever enough to figure out those situations.

    Never say never. It's just a matter of time. Even if some situations are hard to automate, a large percentage of all driving hours (freeway driving, I would think) could be automated much more easily.

    The motivation to reclaim driving time is huge. People spend / waste a fantastic amount of time driving. I couldn't find global figures, but apparently Americans spend over 100 hours per year commuting (not driving in total - just commuting); the total driving figure in Israel is 577 hours per year; and about 40% of mothers in the US spend over 2 hours per day driving. Then there are truck drivers and delivery workers whose annual total must be closer to a couple thousand hours per year (i.e. basically their whole life).

    Dishwashing machines are very popular, and how much time do they actually save, 20 minutes per day? I can't think of anything the average person more, that could be automated as easily, as driving.

  3. too bad for the crew by quitte · · Score: 2, Interesting

    but still I want to see the existing footage now. The teaser clip is pretty cool. Apparently this is not about getting up Pike's Peak but getting up fast. If there are ethical issues showing the helicopter footage at least show the footage from the cars onboard camera that surely exists.
    This is so much more exciting than the stupid soccer bots with their Robocup.

  4. Re:Before jumping to conclusions.... by Rich0 · · Score: 2, Interesting

    I'm not big on the whole "pilot error" thing. Too often it just amounts to blaming somebody for the inevitable.

    Take 200 of the best truck drivers on the planet. Keep them awake for 48 hours straight. Then set them behind the wheels of big trucks at 2AM and tell them to cover a distance of 800 miles at an average speed of 50MPH or greater.

    I can pretty-much guarantee that there will be an accident. Will it be the result of human error - well, sure. However, humans are just another kind of machine. If you took 200 trucks and drove them for 5 years without inspecting their brakes you'd have accidents too - for basically the same kind of reason.

    In this case a helicopter at 14k feet is likely at the edge of its performance envelope. That means that any mistake can get out of hand very quickly. Aircraft aren't safe because pilots don't make mistakes - they're safe because they are engineered and operated in a manner that allows recovery if there is a mistake (while preventing the kinds of mistakes that would be unrecoverable). If you operate an aircraft in a way that is at the boundaries of its operational limits, then sooner or later there will be a crash. Sure, most of the time there won't be one, so it is easy to blame the pilot when the inevitable happens.

  5. The one thing they ARE keeping under wraps by Provocateur · · Score: 3, Interesting

    ...is that the car called to report the accident.

    --
    WARNING: Smartphones have side effects--most of them undocumented.
  6. Ironically... by shrtcircuit · · Score: 5, Interesting

    We were up on Pikes Peak last weekend staffing a charity hike event when the autonomous car itself also crashed, running off the road somewhere. The wrecker they sent up to fetch it also broke down blocking the road, so they had it shut down for a while getting yet another wrecker up the mountain to help relocate the first one, and get the car out of there.

    That thing has some sort of bad omen surrounding it. Everything mechanical around it, including itself, seems to break or crash! I'm amazed nobody has been killed yet, especially with the helo going down on the side of the mountain (that usually ends very badly, so my props go to the pilot for keeping everyone alive).

  7. Insurance costs by nten · · Score: 3, Interesting

    Assuming the autonomous systems actually work most of the time car insurance providers could make a bundle offering discount rates for the feature (only slightly of course, they are evil), and then gradually raising the rates for the lack of the feature. So eventually we won't be able to afford to drive manual vehicles. At least I can read my kindle on the way to work.

    --
    refactor the law, its bloated, confusing and unmaintainable.
  8. Re:So the weak point in the system is...... by Sulphur · · Score: 2, Interesting

    If one has four rotors, wouldn't he use collective pitch on the individual rotors, rather than cyclic pitch or rpm change?