COLMAR-BERG - Pour diminuer le nombre de collision avec les piétons, des tests sont menés sur les systèmes de freinage des véhicules ont été menés ce mercredi au Luxembourg.
Пікірлер: 9
@djodu576 жыл бұрын
Tesla ne semble pas très au point ^^
@DerKatzeSonne6 жыл бұрын
Absolutely unprofessional test with useless results. 1st: Pre-Facelift (= quite old) Tesla with AP1 hardware (if it has Autopilot hardware at all! I don't know if it was optional back then. The newer Teslas since AP2 was introduced, at least have the AEB functionality as standard. All of them.). 2nd: Even if the car had the appropriate (and outdated) hardware: The "dummy" car is made of foam or something, so it doesn't contain any metal structure or anything to reflect the radar waves. At least the AP1 heavily relies on the radar for these emergency breaking situations. So this cannot work. There is a good reason why the NCAP organisation does the AEB tests with a proper dummy. Might be that the AP2 actually relies more on the camera data an therefore might have braked. I don't know. And it doesn't matter as a car should stop for real cars and not for dummy cars. Actually the radar data is very important, for example in fog, the cameras would not be able to see the car in front. 3rd: If this was a really scientific test, they should have mentioned the software version of the Tesla Model S. There have been lots of improvements over the years (and yes, it still gets updates from time to time). Example video for the NCAP setup: kzfaq.info/get/bejne/fNaRmKqB0LDbkaM.html
@bernardhossmoto6 жыл бұрын
My 2016 Model (built in 2015) Mitsubishi Outlander works like the Volvo. No need for "metal hardware", it also brakes in front of brick walls or hay bales. Yes, I have tried it. @katzen_sonne your theory would explain the recent accidents with Teslas crashing into concrete lane dividers. They don't "see" them, because they are not metal.
@DerKatzeSonne6 жыл бұрын
Don't get me wrong. Radar *does* detect other materials than metal. Like a brick wall. Concrete barrier. Maybe hay bales (I don't know for sure). But foam or whatever this is, is one of the worst materials which barely reflect radar waves - maybe also some kind of corner case, where one car triggers and one doesn't (sadly, there is no further information about it; if they released more info, that might proof me wrong, but for now this looks very unprofessional and not like this is proving the Tesla do to something "wrong", if you know some backgrounds about AEB systems and how they work). I don't know, what kind of system your outlander uses for AEB, but seems like the same as the Tesla: Radar. And it would probably not have stopped in the situation in this video. The Volvo obviously uses cameras (maybe additionally a radar? I don't know, but I guess they do, as the cameras alone are not sufficient in my opinion). If it only relies on the cameras, it can detect a car just fine. But hay bales? Do you think they used deep learning to tell it what hey bales are? I highly doubt it. Whatever, let's hope the Volvo doesn't AEB brake when a newspaper with the picture of a car flys right in front of its camera ;) About the recent accident with the concrete barrier: That had a different cause. AEBs normally ignore stationary objects if the car is traveling "fast" (i.e. highway speeds). Otherwise they would break too often, even when they aren't supposed to (more info on this topic and why most likely every other car would have run into the concrete barrier or the fire trucks also: www.wired.com/story/tesla-autopilot-why-crash-radar/ ). As you see, these systems are very complex and there is still a lot of development going on. And I wonder if the current software of the AP2 would maybe even handle that situation better and not crash into a concrete divider anymore (for obvious reasons not because of the radar, but maybe because of the camera vision detecting the street surface and therefor seeing that the car is going off the road; we don't know)? Because it improved a lot since then. I suggest, you watch this video to learn some more about these systems and also a correct crash test dummy: kzfaq.info/get/bejne/huB4gJmBu7rOiKM.html - it's a little bit older already, but many things said in it are still correct. (and after watching this another time, I noticed they say the Volvo and the VW up! have a laser scanner for city emergency breaking; whatever this exactly means, might also be the reason, why the Volvo stopped here and the Tesla didn't - if the 2018 Volvos still have the same system; and still, I don't think that's a real problem in real world situations) And at the end it's important to always keep in mind, that AEB is just a backup system for when the human fails. And it's not guaranteed to work in 100% of the cases.
@caitlingoehrig74176 жыл бұрын
Katze, you should give these resarchers, and all of those in the Automotive testing field more credit. I work for the company who made that target, and it's not an arbitrary design made with arbitrary materials. Most of our targets, including this one, have corner reflectors (small, metal 3D pyramids), and they are placed above license plate level. There is a pocket for a real license plate, as well. There is quilted radar foil that helps the car detect radar signatures, too. Last, but not least, there is LiDAR on the headlights and brake lights on the facing of the slab. The car uses several systems, including the camera, to pick up on obstacles ahead or behind the car. If it doesn't stop, it's not because there wasn't metal in the object. These cars are supposed to avoid anything that the camera sees or senses in the car's path, including humans. For these test purposes, the camera is supposed to pick up on the image it sees, which is the back end of a vehicle. They can't use real vehicles when doing this testing, or they would have to have an astronomical budget for reserach materials. These systems require hours of testing and calibration, before they are ready to be tested with real cars. I've been to an automotive test track, and you don't want to give these engineers real cars/targets to practice with. Start with our targets, and go from there! Moshondata.com 😁
@DerKatzeSonne6 жыл бұрын
Well, that sounds good. Thanks for the clarification. Yep, the website says, that dummy actually has a radar reflective material. Still, this video / demonstration is crap, because they don't give any details. Of course, you wouldn't test with real cars ;) that's for sure. But most other dummies I've seen look different (and a look at your company's website shows that 90% of the dummies you build look like the ones I knew), that's why I questioned this test. Still, many more details to be clarified. And also: You say the dummy has a pocket for a license plate. Have they used it? Cannot clearly tell this from the video.
@caitlingoehrig74176 жыл бұрын
katze_sonne some of them use the real plates, but not all of them. I believe this one has a real plate, because we don't provide graphics for them anymore (used to have digitally printed plates). I think the testing teams that are more strict in their guidelines and parameters do use real plates, but I'm not positive. I think it would be great if these teams gave you information about the target and which parts are picked up by which systems in the car. Most videos tend to focus only on what the camera on the car sees and not what the rest of the car senses.