TMC is an independent, primarily volunteer organization that relies on ad revenue to cover its operating costs. Please consider whitelisting TMC on your ad blocker and becoming a Supporting Member. For more info: Support TMC

AutoPilot Crash today-Tesla response less than stellar?

Discussion in 'Model S: Driving Dynamics' started by carnutfan, Nov 7, 2016.

  1. Az_Rael

    Az_Rael Supporting Member

    Joined:
    Jan 26, 2016
    Messages:
    5,271
    Location:
    Palmdale, CA
    After my Autopilot 1.0 test drive experience with a Tesla sales person, I am not surprised at the instructions you were given. Mine had me engage AP in stop and go traffic with TACC set at 65 and a large gap to the car in front. That caused the car to accelerate towards the stopped car in front of me, and every time I hit the brakes to stop the car (because I dont like accelerating towards stopped cars) the sales person told me to not brake. I guess I was supposed to trust the car. (Which probably would have been fine, except when its not, as in your case)

    I came away perfectly happy to order a AP 0 (no AP) CPO after that.
     
  2. MP3Mike

    MP3Mike Well-Known Member

    Joined:
    Feb 1, 2016
    Messages:
    11,033
    Location:
    Oregon
    Nope, there is no setting for AEB other than on/off. The early/standard/late is for the FCW, Forward Collisions Warning, feature which is separate from AEB.
     
  3. reynirb

    reynirb Member

    Joined:
    Oct 16, 2016
    Messages:
    118
    Location:
    Coral Springs, FL
    In that case it is an engineering flaw in the radar implementation or the radar not working as designed. Most early autonomous systems being implement by other manufacturers are using radars for pre-collission emergency braking, which would be one of the most important features of the system.

    If a system can't react to stationary vehicles in your path, it would be practically useless in the real world.
     
    • Like x 1
    • Disagree x 1
  4. reynirb

    reynirb Member

    Joined:
    Oct 16, 2016
    Messages:
    118
    Location:
    Coral Springs, FL
    Actually, that is not true. The radar signal scatter return is based on the size of the object and its reflected angle, a larger object will have a stronger reflected signal than a smaller one. A radar would never confuse a soda can with a car.

    Same thing with an overhead sign, unless you are driving up a hill towards an overhead sign, the radar would pass underneath it. If you are pointed at an angle to the sign, most of the energy is scatted back at an angle, vs an object like a car would scatter back towards the source.

    "Because of this, the standard approach for car radar is to start by ignoring anything that isn't moving"

    Yet again, you are incorrect. A radar tracking an object at highway speed in front of you sees an object traveling the same speed as you are as stationary. What is why police radars have to be tied to the cars speedometer or an external source like GPS, it has to have a reference or a police radar measuring the car in front of would be going at close to 0 miles per hour.

    The problem with the system is not the radar, it is going to be with the camera. Cameras can track and measure the velocity of objects moving across them, they have a very hard time figuring out if an object coming towards them is stationary or moving, and filter out which objects are supposed to be stationary like signs and which are supposed to be moving like other cars.
     
    • Disagree x 6
  5. Todd Burch

    Todd Burch Voltage makes me tingle.

    Joined:
    Nov 3, 2009
    Messages:
    6,742
    Location:
    Smithfield, VA
    Sorry to hear this, but this scenario should be improved in 8.1. v8.0 is about gathering data to determine when to use the radar braking. The radar braking itself doesn't start getting used until v8.1.
     
  6. GlmnAlyAirCar

    GlmnAlyAirCar Active Member

    Joined:
    Mar 19, 2015
    Messages:
    1,123
    Location:
    Sanatoga PA
    And it relies on the same technology as TACC/AP (radar/camera) so if it doesn't work in one case it won't work in the other.
     
  7. OzSimon

    OzSimon Member

    Joined:
    Jun 14, 2016
    Messages:
    174
    Location:
    Gold Coast Australia
    The autopilot is absolutely awesome, like the rest of the car. The salesman was a total moron, so just order the car and read the owners manual- you will be happy you did.
     
    • Like x 2
  8. Saghost

    Saghost Well-Known Member

    Joined:
    Oct 9, 2013
    Messages:
    7,798
    Location:
    Delaware
    A soda can bottom makes a beautiful radar corner, returning a much stronger signal than an average object of its size. If I remember right, Elon used it specifically in his discussion on the subject when introducing the whitelist approach. Even phased array radar knows nothing about the size of an object, only the strength of the return - which varies based on the size and geometry of the object - small things with radar corners look much bigger than large things that are convex, or especially flat plates that aren't perpendicular to the source.

    Do you have a source for saying that the standard approach isn't to disregard everything that isn't moving? (as in, everything that appears to be moving towards you at your current speed, as I clarified in the parenthesis in my original post, before you attempted to lecture me on it.) I've read that it is the normal approach in a number of places, and it is in fact the principle reason that pulse-doppler radar technology was developed - to break aircraft out of ground clutter.
     
    • Like x 4
  9. MP3Mike

    MP3Mike Well-Known Member

    Joined:
    Feb 1, 2016
    Messages:
    11,033
    Location:
    Oregon
    Yes he did, from the blog post on using radar for more: "On the other hand, any metal surface with a dish shape is not only reflective, but also amplifies the reflected signal to many times its actual size. A discarded soda can on the road, with its concave bottom facing towards you can appear to be a large and dangerous obstacle, but you would definitely not want to slam on the brakes to avoid it."
     
  10. hacer

    hacer Member

    Joined:
    Apr 13, 2016
    Messages:
    941
    Location:
    Clarksville, MD
    You know just enough about radar to be dangerously wrong about it. Stationary objects, e.g. the Earth and everything on it, are quite large and generally fill much more of the radar field of view than vehicles in traffic (except when they are super close to the front of your car). The vast majority of reflected energy returning to the radar at nearly every range is from these stationary objects - it's called clutter. Doppler radar allows you to filter out this clutter and very reliably see the things that are moving. The radar cross section of objects depends on a whole lot of things, size is only one of them; a well-aimed convex bottom of an aluminium soda can will produce a larger radar cross section than some cars because it acts like a retro-reflector when on axis, while some cars have smoothly curved convex surfaces of composite material that return little energy back to the radar. To some limited extent you can try to sift through the clutter by observing its spatial pattern and how it changes as you close distance on it but that is very crude. If you relied on it for braking the car you would shut the system off because of all the false positives. The spatial resolution (except on the range axis) is very poor for automotive radars which certainly doesn't help. The bottom line is that it is really hard to distinguish dangerous stationary objects that you might hit from all the other stationary objects that you definitely won't hit with radar alone.
     
    • Like x 3
    • Informative x 2
  11. S4WRXTTCS

    S4WRXTTCS Active Member

    Joined:
    May 3, 2015
    Messages:
    4,654
    Location:
    Snohomish, WA
    #51 S4WRXTTCS, Nov 8, 2016
    Last edited: Nov 8, 2016
    The way I see it is any car company that has an adaptive cruise control system needs to have the ability to see a stopped car especially when it's combined with a lane-steering system. If they don't it's inevitable that the driver is going to trust the car, and the car is going to fail.

    Stopping for a stopped object seems so easy, but it's a really tricky thing to implement while also not getting false positive.

    One of the best AEB systems on the market is Subaru's Eyesight. It fulfills both objectives you listed (at up to around 30mph speed difference), but there are complaints about false positives. It's also really hard to actually find it on their cars (like try finding a CPO Crosstrek with it on it).
     
  12. bonaire

    bonaire Active Member

    Joined:
    Aug 24, 2013
    Messages:
    2,459
    Location:
    USA
    Question for AP 2.0 planning. How will it handle potholes? Specifically, someone's eyes pick up on nasty potholes which they will try to avoid by going a little left or right of it. Will a pothole factor into decision making by the radar and/or can a camera pick it up?
     
  13. Saghost

    Saghost Well-Known Member

    Joined:
    Oct 9, 2013
    Messages:
    7,798
    Location:
    Delaware
    That's a good question, with no clear answer yet.

    Both the cameras and the radar should be able to see it, though the radar return will be buried in the rest of the clutter unless the car goes looking for it specifically (presumably based on the cameras seeing it.)

    One more thing Tesla will have to sort out before level 5 operation is practical.
     
  14. bonaire

    bonaire Active Member

    Joined:
    Aug 24, 2013
    Messages:
    2,459
    Location:
    USA
    In the rain, water-filed pothole that human eyes may not be able to see but can "think is possibly there" and avoid. Also, once you see it on a dry day, you remember that pothole and avoid it on your next commute. AP 2.0 may not have such memory in the rain. Edge case conditions, sure, but my brain thinks like this and you're in Delaware and know about winter-tie havoc in terms of potholes.
     
  15. scottm

    scottm Version 9 software sufferer

    Joined:
    Jun 13, 2014
    Messages:
    3,070
    Location:
    Canada
    #55 scottm, Nov 9, 2016
    Last edited: Nov 9, 2016
    Here's the short story: your friend rear ended a car on a clear day. Judgement: your friend is at fault. No matter what kind of vehicle was being used or its capabilities or lack thereof. I think people have woken up now about this being the reality of autopilot cars in today's world.

    There's one or two or more tickets for that, curious which ones did he get ?
    - not retaining control of vehicle at all times
    - following too close
    - unsafe driving
    - distracted driving
    - traveling faster than conditions allow
    - etc..

    The occupants and owner of car that got hit are going to turn to the driver and Tesla company to pay damages, as soon as they learn this was a test drive. Driver coughed up his ID and drivers licence and insurance policy, right?

    Being the driver, he is guilty of wrecking a Tesla demo vehicle. There's something you sign when going out on a test drive, so that kicks into effect. Tesla will follow that and probably pursue his insurance for the damages to their car. (Maybe he can buy the written-off car and you can buy it as spare parts, just an idea.) If his insurance doesn't cover it (no insurance) he might be personally liable for covering the cost. Ouch! Having to buy a wrecked car is no fun.

    What happens next is the interesting part... the salesman. What did he say during the drive? Was there a dash cam or other voice recording going on throughout this event to catch what was said, as a "suggestion" (your words) not to brake. What exactly was said? Was it a command to not brake, or a suggestion to allow the car an opportunity to brake? Can you discern clearly on the recording who said what? If there's no recording, it gets more interesting.

    If not, is the third party witness (you) credible and unbiased enough to be believed by judge? Hmm... versus word of salesperson. "Understanding and apologetic" BUT did he admit or deny saying "don't brake" with any other witnesses around? What was his written statement for the crash. Does he admit to interfering with your driving choices by suggesting the car would operate in a way he was purposefully demonstrating?

    Regarding your cancellation: I would expect it to be handled like any other request to cancel within a permissible cancellation period. As per policy. You request, they cancel, and handle any deposit return or loss, per those terms you agreed to.

    The fact that a Tesla employee handling your cancellation is unaware of your test drive experience IS A GOOD THING. They should NOT KNOW about your experience because they don't need to know about that information to handle your request, frankly it is too much information. Maybe even protected under privacy and access to information acts. However, they searched / found /' saw / read that information because you asked them to. But has no bearing on handling your cancellation, they should follow that policy as they would for anybody else and I hope that's what happened.

    The crispness of events that followed with email contacts and updates and letters... I say are typical of Tesla (or any other similar company) for a routine transaction. Things are typically handled more crisply on the "sales and confirmation of order" side... but when a potential customer is decidedly retreating from a sale, the energy and enthusiasm to delight would drop sharply - understandable.
     
    • Informative x 3
    • Like x 1
    • Disagree x 1
  16. bonaire

    bonaire Active Member

    Joined:
    Aug 24, 2013
    Messages:
    2,459
    Location:
    USA
    They're trying to sell auto-pilot type features to an uninformed mass of people and encouraging them (during test drives) to do what may be illegal operation of a vehicle or at least causing confusion in an already-trained human brain. This has to be managed with a very clear eye of liability of risk on the part of all parties - and definitely treat AP as a figurative "loaded-gun" in the hands of an untrained customer. I have to believe that the OP's friend now is liable for the accident, any pending litigation by the one they hit and more costs of rising insurance rates and so on. When the one that was hit realizes he was hit by "auto-pilot managed car owned by a billionaire-run company" their lawyer may be happy to serve them in their best interest. OP's friend certainly is in the hot seat.
     
  17. scottm

    scottm Version 9 software sufferer

    Joined:
    Jun 13, 2014
    Messages:
    3,070
    Location:
    Canada
    #57 scottm, Nov 9, 2016
    Last edited: Nov 9, 2016
    Really, many lessons learned are...
    • understand your insurance position going into test drives, are you covered, are they covered if you don't have insurance
    • only test vehicles you can realistically afford, you might be buying it that day
    • read the thing you sign for a test drive, if you don't like it don't sign it or don't drive
    • stick a dashcam on the windshield for the test drive, audio capture on, inform other occupants, let it roll for the whole time you're behind the wheel
    • learn the standard controls of the car in a safe area with no traffic
    • know where your brake pedal is and get the feel for the brake pedal, stop hard (panic!) somewhere with no traffic, learn your stopping distance... warn your occupants prior, tell them you're testing the ABS
    • don't experiment with any extraneous features on your very first drive. Steering, go, stop, directional changes, mirrors... pay attention to traffic and drive. That's it.
    • ask for an overnight loan of the demo car to try your regular commute cycle, and experiment with features as you gain familiarity
    • follow the letter of the law when driving a test vehicle - from your own good head - don't take advice from passengers
    • don't talk or be distracted by salesmen, your phone, or anybody or anything else in the car while driving. Even the radio / media is turned off.
     
    • Informative x 2
    • Helpful x 1
    • Like x 1
  18. sdorn

    sdorn Director of Awesome

    Joined:
    Oct 9, 2016
    Messages:
    921
    Location:
    Georgia
    During my test drive the salesman had me engage TACC on a divided highway with stop lights. Twice as I approached cars stopped at a light the car began to brake and then stopped braking, forcing me to aggressively brake in order to avoid hitting the stopped car in front of me. The salesman's explanation was that it was a new car that hadn't been driven enough yet for the TACC to work properly. He said it typically takes 2 - 3 weeks of normal driving before TACC, auto-steer and self-parking accumulate enough experience to start working properly.

    This was on a car with AP1 and 8.0 firmware.
     
  19. ccharleb

    ccharleb Member

    Joined:
    Dec 19, 2014
    Messages:
    142
    Location:
    Montreal (QC), Canada
    This!!!

    When I'm following a vehicle with TACC or AutoSteer engaged, I always confirm that the car ahead of me is displayed on my Instrument Cluster, in white. If the vehicle is shown, TACC will maintain proper distance. If not shown, TACC is NOT seeing that vehicle! You must be prepare to slow down yourself.

    Always scan that instrument cluster.... it's full of great feedback from your sensors!

    Christian
     
    • Like x 2
    • Informative x 1
  20. sdorn

    sdorn Director of Awesome

    Joined:
    Oct 9, 2016
    Messages:
    921
    Location:
    Georgia
    Salesmen don't tell you about that sort of thing during test drives, at least mine didn't.
     

Share This Page

  • About Us

    Formed in 2006, Tesla Motors Club (TMC) was the first independent online Tesla community. Today it remains the largest and most dynamic community of Tesla enthusiasts. Learn more.
  • Do you value your experience at TMC? Consider becoming a Supporting Member of Tesla Motors Club. As a thank you for your contribution, you'll get nearly no ads in the Community and Groups sections. Additional perks are available depending on the level of contribution. Please visit the Account Upgrades page for more details.


    SUPPORT TMC