Featured Tesla removed Radar, now Sonar too.

Discussion in 'Prius, Hybrid, EV and Alt-Fuel News' started by dbstoo, Oct 18, 2022.

  1. dbstoo

    dbstoo Senior Member

    Joined:
    Oct 7, 2012
    1,178
    620
    0
    Location:
    Near Silicon Valley
    Vehicle:
    2017 Prius Prime
    Model:
    Prime Advanced
    When radar chips became sparse, Tesla suddenly stopped using radar in their cars, citing the difficulty of merging the logic of video based logic with data from a radar feed. Over the course of a few months they were able to replace much of the auto-pilot functions with video based environmental awareness. Now they are dropping the ultrasound sonar sensors that are used to locate nearby obstructions, including pedestrians. It looks like they are citing cost savings.

    I find this change to be odd since it comes at a time when most car manufacturers are improving their ability to detect and avoid accidents by using more sensors and AI.

    The link below has a video that shows that there are some really big blind spots in the car's camera field of vision. Most alarming is that there are areas in front of the car that would not see a child playing.

    Tesla saves an estimated $114 per car by removing USS
     
  2. bisco

    bisco cookie crumbler

    Joined:
    May 11, 2005
    102,879
    46,684
    0
    Location:
    boston
    Vehicle:
    2012 Prius Plug-in
    Model:
    Plug-in Base
    sometimes i wonder if self driving cars are just a fantasy
     
  3. Trollbait

    Trollbait It's a D&D thing

    Joined:
    Feb 7, 2006
    20,149
    10,264
    0
    Location:
    eastern Pennsylvania
    Vehicle:
    Other Non-Hybrid
    The radars weren't needed for most of these driver aid and safety functions. Subaru's Eyesight is one of the more highly rated systems, and it doesn't use radar for the forward view.

    Getting rid of the sonar just seems like a bad idea though. Tesla's seem to do a good job of using memory of stationary object locations when they pass into a blind spot, but there are things that can move once out of sight.

    Tesla hasn't been expanding service to keep up with sales. Getting rid of sonar on new cars saves production costs. It also is one less thing to burden service with by eliminating the need for calibration after body work. Tesla will be turning off the sonars already on cars at some point in the near future.
     
  4. 3PriusMike

    3PriusMike Prius owner since 2000, Tesla M3 2018

    Joined:
    Jun 21, 2009
    2,826
    2,186
    0
    Location:
    Silicon Valley
    Vehicle:
    2012 Prius Plug-in
    Model:
    Plug-in Base
    I don't think the ultrasonics sensors are used to detect pedestrians, specifically. Do you have a link for this?
    The effective range is about 30-36 inches. If you are traveling at any significant speed at all if you just detect something you have no time to stop. At 2 m/h you go ~3 ft/sec

    As you creep forward, for example, near a parking structure wall, it starts to beep and show your distance about 30 inches away. The useful distance of sound based sensors is limited by the frequency, sample rate and speed of sound so they are not as useful as light or radar because in order to get a longer range they have to reduce the sample rate.

    As I understand it the ultrasonic sensors are used for parking and staying away from freeway barriers, etc. It seems that the plan is for these objects to be visually seen and held in a persistent memory to obtain the same functionality.
    They should have gotten this new capability working before announcing the change, IMO.
    But the Nissan Leaf has had a bird's eye 360 view display with fewer cameras for over 7 years.

    Mike
     
  5. hill

    hill High Fiber Member

    Joined:
    Jun 23, 2005
    18,574
    7,425
    54
    Location:
    Montana & Nashville, TN
    Vehicle:
    2004 Prius
    Model:
    IV
    multiple accident threat machine logic signals ... often confuse AI - at which point it may cause phantom breaking. Hard to guess what motive tesla sees as best, beyond price. I sure did enjoy using their mobile eye, until the 2 companies parted ways.
    .
     
  6. dbstoo

    dbstoo Senior Member

    Joined:
    Oct 7, 2012
    1,178
    620
    0
    Location:
    Near Silicon Valley
    Vehicle:
    2017 Prius Prime
    Model:
    Prime Advanced
    No, I don't have a link that says that ultrasonics sensors are used to detect pedestrians, specifically. I gave up trying to nail them down on specifics when Musk was tweeting about how superior his business model was, allowing changes in manufacturing from day to day without having to remain backward compatible.

    But I've seen MY car detect people walking in front of and behind my car when exiting a parking space. I'm pretty sure that's done by ultrasonic sensors. My radar, on the other hand, looks for cross traffic approaching from a distance and warns me in plenty of time even when I'm parked between two pickup trucks, SUVs, etc.

    Relying on visual cues is probably easier to handle when you have an AI that's scrubbing video captures, but it seems to suck big time when you have low contrast scenes or when detecting things that are stationary and uncommon. The AI can't reliably interpret scenes that were never presented to it, right?

    The problem I see with visual feeds for object detection is that the sample rate is relatively slow when you have to wait long enough for differences to show up between two frames. Only then can you correct for jitter and vibration and then finally apply algorithms to determine the distance from the parallax. How quickly can an optical system do that depending only on cameras? Can it plot the path of a car approaching from the left at 60 miles per hour before it enters an intersection? Judging from some of the YouTube videos that I've watched, the technique Tesla uses is often uncertain about oncoming traffic when the car has to turn left to merge into traffic. By uncertain I mean that it waits an extraordinarily long time for traffic to clear.
     
  7. 3PriusMike

    3PriusMike Prius owner since 2000, Tesla M3 2018

    Joined:
    Jun 21, 2009
    2,826
    2,186
    0
    Location:
    Silicon Valley
    Vehicle:
    2012 Prius Plug-in
    Model:
    Plug-in Base
    Are they less than ~3 feet? Does it also detect them when they are 5 or 10 feet away or more?

    The AI doesn't need to reliably detect what it is seeing, just that there is an object of some kind there. Every possible type of object doesn't need to be presented to it during training. If you want it to know is it a cat or a dog, then you'd need to show it both. Overwise it might say everything is a dog that has 4 legs and a tail even if it sees a cat.

    The Tesla cameras operate at 36 fps. Most cameras operate at 30 fps. The HW is designed to run the inference fast enough to keep up with that frame rate...i.e. about 25 ms. It takes the human body about 50 ms just to convert a decision by the brain to act before a muscle begins to move. It takes another 200-300 ms for your foot to fully depress the brake pedal in an emergency stop after you decided to stop. It is a total of about 750 ms for a human to see, judge and react.

    source:
    Stopping time and distance | Alberta.ca

    Yes, it is being over cautious, perhaps, like someone with a learners permit.

    Mike
     
  8. dbstoo

    dbstoo Senior Member

    Joined:
    Oct 7, 2012
    1,178
    620
    0
    Location:
    Near Silicon Valley
    Vehicle:
    2017 Prius Prime
    Model:
    Prime Advanced
    I asked: Can it plot the path of a car approaching from the left at 60 miles per hour before it enters an intersection? Judging from some of the YouTube videos that I've watched, the technique Tesla uses is often uncertain about oncoming traffic when the car has to turn left to merge into traffic. By uncertain I mean that it waits an extraordinarily long time for traffic to clear.

    Mike was nice enough to reply:
    Yes, it is being over cautious, perhaps, like someone with a learners permit.

    Do you know whether this over cautious approach due to inability to determine the speed of the oncoming car? I suspect that it needs the cars to be in a certain number of frames over a certain period of time before the deltas can be accurately calculated.

    I find that accurately assessing the speed of a car that is approaching head on is one of the challenges that many people do poorly. Guess wrong and you are liable to be in an accident. Quickly determining the speed of an approaching object is one of the major advantages that radar and lidar have.
     
  9. 3PriusMike

    3PriusMike Prius owner since 2000, Tesla M3 2018

    Joined:
    Jun 21, 2009
    2,826
    2,186
    0
    Location:
    Silicon Valley
    Vehicle:
    2012 Prius Plug-in
    Model:
    Plug-in Base
    There have been numerous techniques used to approximate the speed of cars in traffic using only video for more than a decade so I think any method used by Tesla would be equal of better than these.

    For example, here is an iPhone app that does it

    https://road.cc/content/news/new-app-allow-public-submit-evidence-speeding-291501

    Mike
     
  10. dbstoo

    dbstoo Senior Member

    Joined:
    Oct 7, 2012
    1,178
    620
    0
    Location:
    Near Silicon Valley
    Vehicle:
    2017 Prius Prime
    Model:
    Prime Advanced
    From speedcam anywhere web site, the requirements for it to work are pretty restrictive:

    Speedcam Anywhere needs to be able to get a clear view of the whole vehicle passing across the screen. It needs to have a clear view of the wheels, and to get an accurate measurement, a clear view of the numberplate as the vehicle passes. Aim the camera at about 45 degrees across the road for best results. The phone must be held still for 2 seconds whilst the capture takes place. Do not pan the camera to follow the vehicle. Practice in Basic Mode before using more credits in Pro Mode.

    Lower spec phone cameras, rain, bad light and being further away from the vehicle will reduce the likelihood of the capture succeeding.


    After all that, it has to send the video to a server farm where the actual calculations take place. I could probably calculate the speed quicker just based on known dimensions and how long it takes to cross the field of view.

    But this does not appear to work for approaching traffic, nor for situations where the car with the camera is moving.
     
  11. 3PriusMike

    3PriusMike Prius owner since 2000, Tesla M3 2018

    Joined:
    Jun 21, 2009
    2,826
    2,186
    0
    Location:
    Silicon Valley
    Vehicle:
    2012 Prius Plug-in
    Model:
    Plug-in Base
    I wasn't pointing to this particular app as a perfect example. I was just pointing out that there has been lots of prior academic research into computing car speed from videos, some of which have been put into products over the years. The idea of doing this (car speed from video) isn't something Tesla just thought up...the research and incrementally better implementations go back a decade or more.

    Mike
     
Loading...