6 ms·

States just need to start mandating LIDAR for any level of autonomous certification, in the same way we finally figured out seatbelts were the minimum level of

• njarboe 6y ago
  Mandating how a certain outcome should be achieved generally is not the best (efficient, quickest, cheapest, etc.) way in getting to that outcome. Autonomous certification should have its safety requirements for each level and require the vehicles meet that standard. Autonomous vehicles have the potential to reduce deaths and injuries dramatically and increasing the time to get there will cause thousands to millions of more deaths and an order of magnitude more injuries, not to mention property damage.
  • nexuist 6y ago
    This isn't necessarily that strict of a regulation. They could require the sensor be on board but not actually regulate any of its usage. It could be a scenario like the Model 3's rear view mirror camera, where it wasn't used at all for the first two or so years and then later got a software update that started interacting with it. The important part is that the LIDAR unit is physically wired in to the system so that they can start using it in software whenever they feel like it's necessary.
  • tshaddox 6y ago
    I don't know. In general I think you're going to need some balance between requiring certain outcomes and requiring certain specific implementation details. It might be possible to make an automobile out of sufficient thick and squishy parts that it could pass reasonable crash safety tests without having a seat belt, but it's still not ridiculous for regulations to require that automobiles have seat belts. (I have no idea where lidar for self-driving cars falls within that balance, although I can't help but notice that human drivers do not utilize lidar.)
    • DenseComet 6y ago
      Well, lidar is just a stand-in for the depth perception humans already do. It is absolutely valid for regulators to require that the systems on an autonomous car do depth perception, and provide a standard to prove that the system works well. I suspect that right now, ML based approaches to depth perception would not pass the test to standards we would like, in practice requiring lidar. But, this would also provide an option to switch to an ML based gets good enough.
      • acadien 6y ago
        This is a common misperception. You cannot estimate the shape of simple things in low light. Even humans are susceptible to optical illusions and are effectively blind during certain times of day (try driving in the direction of the sun during sun rise/set). Lidar is not susceptible to any of these things. Similarly ML will never solve low light or direct light situations, that is a physical limitation of cameras.
      • tshaddox 6y ago
        There are popular cars in the US that use two cameras for their self-driving functionality. Kinda like humans.
        • lttlrck 6y ago
          Those cameras are fixed whereas eyeballs have multiple degrees of movement via the eye socket and skull mounting apparatus allowing very sensitive parallax measurements. It's really not the same at all.
  • seanhunter 6y ago
    Autonomous vehicles do of course have the potential to reduce deaths but there are lots of examples of this being the preferred method of regulation in the auto industry. For example if you want to sell a motorcycle in the US it has to have a kill switch in a specific place on the right handlebar. You don't (as a manufacturer) get to innovate about this because it's a safety feature and a first responder needs to be able to locate the kill switch so it always has to be in the same exact place so they can always find it. If you want to sell a car in Europe (I believe) it has to have seatbelts, ABS and a catalytic converter that meets a particular specification etc etc.
• aaaxyz 6y ago
  Not necessarily mandating LIDAR but at least mandating functional redundancy for self-driving / driving assistance. I'm guessing that won't happen until the deaths start piling up though.
• contravariant 6y ago
  Is LIDAR not susceptible to this? I guess it might be more robust against random hiccoughs (though I have no basis for this whatsoever), but is it also protected against malice?
  • ianmobbs 6y ago
    LIDAR maps out the world around you in three dimensions, so with that data, there's no way it would mistake "a 'Stop' sign hidden within a fast food commercial [...] only flashing on-screen for a fraction of a second" (article) for an actual stop sign
    • brianwawok 6y ago
      But it would fall for a cardboard stop sign on a stick held up by a troublemaker.
      • dwohnitmok 6y ago
        As would a human given a sufficiently official-looking stop sign. In fact I think it's actually impossible for an automated car or human without figuring out the "troublemaker"'s identity whether the stop sign is fake or not. And it's not always possible to pin down someone's identity (if the person is wearing a brightly colored construction vest what do you do?).
        • couchand 6y ago
          There's a lot of construction in our neighborhood right now which made me start thinking about that possibility. It kind of sounds like something out of a heist movie (or taken to the extreme, Die Hard With a Vengance?), but really, who would know and care enough to ask the questions to notice something like that?
      • noodlesoups 6y ago
        That requires a whole different kind of effort. These hacks can be done remotely from the other side of the world and leave little trace
• Diederich 6y ago
  Can LIDAR reliably detect stop signs? I think it would have to be based on the shape, and LIDAR loses resolution pretty quickly over distance.
  • airstrike 6y ago
    You could always map the stop signs in advance
    • kurthr 6y ago
      until they move, get removed, or are run over...
    • dreamcompiler 6y ago
      Which would fail in the case of construction workers holding up stop signs or school buses deploying stop signs, to cite just two examples.
      • nexuist 6y ago
        It's actually comical in a way that we possess the technology to build robots that can operate heavy machinery and practically interpret the world the same way we do, but we still have to worry about people holding up plastic signs. It makes me think that the first real wide scale self driving vehicle deployment will have to come from a nation state rather than a corporation; the easiest way to solve all of these problems is by putting heavy investments into smart road infrastructure such that the car depends on sensors embedded in the signs or the asphalt rather than any camera feed. A school bus could just have a transmitter to indicate whether its lights are on or not; construction workers could be given road signs that transmit similar signals or better yet use a portal to construct a virtual geofence around wherever the construction is happening before it begins. These are all easily solvable problems in today's world; what is missing is the political capital to invest in these solutions despite the insane ROI.
        • dreamcompiler 6y ago
• GuB-42 6y ago
  Tesla cars are far from being autonomous. Autopilot (including the so called "full self driving") is level 2, meaning that the driver has to pay attention at all times. From a driver responsibility point of view, it is a fancy cruise control, there is no question about autonomous certification because the car isn't autonomous. If one day, Tesla thinks it is OK for drivers not to pay attention and wants to make it legal, that's when the question about what kind of tech is reasonable will be asked. But for now, it looks like we are far from it. There is a huge gap between "works well most of the time" and full autonomy.
  • thelean12 6y ago
    I've said it for a long time but: Tesla implementing level 2 is dangerous and irresponsible. Calling it "autopilot" is even worse (and has been ruled as misleading in at least German courts). If a human has to sit there and monitor the car without interacting with it, then we shouldn't allow it on the roads. That means skipping level 2 and probably level 3. Go the Waymo route and go straight to full self driving cars.
  • nexuist 6y ago
    This day came and passed years ago. It's evident from their marketing that there is no attempt to discern fancy cruise control from "zomfg your car will drive you around in a year!!" Whether this is false advertising or even just moral is up in the air, but Tesla is making no attempts at educating consumers about the limits of their autonomy other than the bare minimum (yeah you technically have to keep both hands on the wheel, but that mechanism is easily defeated and you can even buy clips on Amazon that do it for you). I'm pretty sure Elon even liked a tweet about a Pornhub video where someone banged a model on the highway with autopilot. Which is an excellent day for PR and a disastrous day for legal. The point being, sure, academically they are a far way away, but they are trying to convince consumers Level 5 is just around the corner.
    • tshaddox 6y ago
      I'm coming around to agreeing with you about the deceptiveness of Tesla's marketing of Autopilot. I recently bought a new car from a non-luxury brand that's quite popular in the US, and a few days after buying it I was a bit surprised to realize that its "dynamic cruise control" and "lane keep assist" features amount to roughly the exact same thing as Tesla's current Autopilot offerings. I am still shocked at how undersold these features were, both by the dealers and the car's literature. I was driving for several days before I realized that, in addition the the dynamic cruise control I was familiar with, a simple button press could get the car to steer itself to stay in its lane.
• dawnerd 6y ago
  Why? Cameras can work just as well as your vision. Humans manage just fine and don't have LIDAR. Just because it causes the cars to slow down now doesn't mean it will in the future.
  • sli 6y ago
    It's a little strange to make a direct comparison between human cognition and computer vision given that the latter lacks the true ability to understand what it sees. You can train the computer for these edge cases, but humans largely do not have them in the first place because we're capable of formulating highly accurate understandings of what we see. You can of course pick out discrete examples of humans failing tasks like these, but entire classes of computers routinely fail at things like this.
  • roughly 6y ago
    Your vision is backed by a neural network with around 80 billion neurons implementing a general learning and intelligence model we don't have the first clue how to replicate. We're a bit further away from "Cameras working just as well as your vision" than I think you're anticipating.
    • throwaway2048 6y ago
      Also your eyes are far better at quickly adapting to things like wild changes in brightness and contrast than any camera
    • tshaddox 6y ago
      > We're a bit further away from "Cameras working just as well as your vision" than I think you're anticipating. Well surely that's what we should be testing. Is a self-driving car system safer than a human, or not? You're correct that self-driving car systems lack a "general learning and intelligence model," but they also have several obvious advantages over humans, like potentially better-positioned cameras, a more reliable "attention" system, faster reaction times, potentially more "experience" with different road conditions and scenarios, etc.
  • 6y ago
• gruturo 6y ago
  While I agree with you on mandating it, a LIDAR "only" gives you a 3D point cloud. Great for detecting if an object is in your way - but it cannot read a stop sign, you still need a camera system for that - so they still have to harden it against this hack (and who knows how many to come).
  • tshaddox 6y ago
    Wouldn't that "hack" just be as easy as sticking an actual stop sign somewhere it's not supposed to be, or removing a real stop sign? That would obviously fool a human driver just as much, and it has actually happened, and killed people.
  • bonestamp2 6y ago
    > so they still have to harden it against this hack True, but LIDAR should make that detection more robust. For example, if a stop sign is on the same plane as the object that surrounds it, that's a strong signal that it is not a real stop sign but rather a projected or displayed one. That's much harder to detect with computer vision alone, especially if they're dealing with hackers who can test it with a real car and continually iterate on it to find different ways to trick it.
    • nitrogen 6y ago
      For example, if a stop sign is on the same plane as the object that surrounds it, that's a strong signal that it is not a real stop sign but rather a projected or displayed one. It's conceivable that in denser or older areas, signs are sometimes placed on structures because there isn't room for a pole.
      • bonestamp2 6y ago
        Agreed. I didn't say it's definite, I said it's a strong signal. Even strong signals need to be weighed against other available data. Apple Maps shows stop signs and traffic lights on the map. A database like that would be another strong signal, but also not always accurate.
• glaucon 6y ago
  Not being sarcastic but how would LIDAR help here ? Maybe .... before acting on a Stop sign sensed by the cameras the control system would confirm, via LIDAR, there is a post with a plate mounted on it in a position which corresponds to where the camera thought it saw a Stop sign ? Can LIDAR do that level of sensing, ie unambiguously find a metal post/plate out of all the other stuff which might be in front of the car ? And assuming my guess about how LIDAR would help is correct what happens when an obstruction between the car and the Stop sign means that LIDAR can't 'see' the post but the cameras can see the sign ?