Tesla crash ... with no driver?

  • SUPPORT THE SITE AND ENJOY A PREMIUM EXPERIENCE
    Welcome to Tesla Owners Online, five years young! For a low subscription fee, you will receive access to more features on the site. We now offer yearly memberships too! For more information visit this page:
    https://teslaownersonline.com/account/upgrades

    FOLLOW US ON TWITTER

    SUBSCRIBE TO OUR YOUTUBE CHANNEL
    Did you know we have a YouTube channel that's all about Tesla? Lots of Tesla information, fun, vlogs, product reviews, and a weekly Tesla Owners Online Podcast as well!

help_balance_thegrid_with_evdotenergy

Madmolecule

Top-Contributor
Ludicrous Level Supporter
Joined
Oct 7, 2018
Messages
829
Location
Duluth, GA
Country
Country
Tesla Owner
Model 3
( And why are we talking about autonomous cars? We don't have those yet. We have driver aids - and ALL of those aids are still ßeta (and any real definition of ßeta means "feature complete". with bugs to be resolved - and we're not even close to "feature complete" so, in reality, what we have now is αlpha code. )

We are only talking about autonomous/full self driving because that is what I was sold years ago. I agree to call it Beta you would at least think they would have to defined the capabilities of the software, And at least release the specs.
All the data that Tesla touts that proves their vehicles are safer when an auto pilot then without it’s somewhat misleading. Basically the way they should present the data is that the combination of the driver fully in control along with the automated auto pilot is safer than the driver alone. I don’t know that they have any auto-pilot data that is representative.
What will the city streets functionality be, we all have Fantasy Uncertainty and Doubt about his capabilities But none of us know what it will actually do. Again this is something I paid for almost 2 years ago.
I’ve been laughing at the roadster buzz from a little plaque they put up at a museum. Not a video of even a car doing 0 to 60 in 1.1 seconds. Instantly all the bloggers come out and say it’s been confirmed, the roadster I can really do 1.1 seconds. Nothing has been confirmed they put up a plaque at a museum, and Elon said it will have fart thrusters via tweet. They have taken the corporate mentality of a plaque and a promise. They brought it into modern times by adding a follow up tweet by Elon. Yeah they don’t need a PR department. They get all the marketing they need from existing customers, and by giving away free roadsters fantasies

Also removing radar was a terrible move. It’s one thing if you never had it. The first time a Tesla runs over a kid without a radar, I would imagine the Defense for the victim would certainly ask is there anyway this tragedy could’ve been prevented if there was radar in the car. It’s one thing I have never had radar and be able to honestly answer we don’t know but we don’t think so. It’s another thing to have millions of miles of radar data showing that in fact it might’ve helped even in the slightest way to save the kids life. You would think if there was a supply chain issue tesla would’ve recalled early model teslas they did not purchase full self driving and Rob the radar is out of them. I still think it is a patent infringement issue where Tesla wants to be the only vision only Supplier and if they can make that work they would have a lock on it. It’s only my belief but I’m trying to find something that makes sense. But they did spend money on aero wheels, and cut weight by reducing insulation. I’ve got extended range but terrible road noise, which sadly sometimes my immersive sound can just barely overcome.

Elon Electrify Cuba
 
Last edited:

Klaus-rf

Top-Contributor
Joined
Mar 6, 2019
Messages
837
Location
Southern (Oro Valley), AZ
Country
Country
Tesla Owner
Model 3
All the data that Tesla touts that proves their vehicles are safer when an auto pilot then without it’s somewhat misleading. Basically the way they should present the data is that the combination of the driver fully in control along with the automated auto pilot is safer than the driver alone. I don’t know that they have any auto-pilot data that is representative.
You bring up a very good point here - there is ZERO data to show AP/FSD all by itself (since it doesn't exist) is safer than human-only drivers. Since there is no non-human-driver data to compare it with.

Reminds me of my favorite t-shirt logo:

"There are two kinds of people in the world: Those that can extrapolate from incomplete data"
 
  • Like
Reactions: Madmolecule

M3OC Rules

Top-Contributor
TOO Supporting Member
Joined
Nov 17, 2016
Messages
974
Location
Minneapolis, MN
Country
Country
Tesla Owner
Model 3
Liability is something that needs to be worked out but I don't believe that's a major limiting factor. Tesla providing insurance will go a long way to alleviate that. But other insurance companies will probably want in on the game so they will figure it out. Look at Waymo. Liability is not their biggest problem. Its capability and scaling. It doesn't run people over but it might get stuck somewhere. Basically engineering problems. I think once they get to the point where its competent to run on its own the number of overall accidents will be less but more importantly the number of fatal accidents caused by FSD will be much less than humans.
 
  • Agree
Reactions: DocScott

Tesla4Me!

Member
Joined
Jul 29, 2017
Messages
23
Location
Germantown, MD
Country
Country
Tesla Owner
Model 3
We are only talking about autonomous/full self driving because that is what I was sold years ago. I agree to call it Beta you would at least think they would have to defined the capabilities of the software, And at least release the specs.
All the data that Tesla touts that proves their vehicles are safer when an auto pilot then without it’s somewhat misleading. Basically the way they should present the data is that the combination of the driver fully in control along with the automated auto pilot is safer than the driver alone. I don’t know that they have any auto-pilot data that is representative.
What will the city streets functionality be, we all have Fantasy Uncertainty and Doubt about his capabilities But none of us know what it will actually do. Again this is something I paid for almost 2 years ago.
I’ve been laughing at the roadster buzz from a little plaque they put up at a museum. Not a video of even a car doing 0 to 60 in 1.1 seconds. Instantly all the bloggers come out and say it’s been confirmed, the roadster I can really do 1.1 seconds. Nothing has been confirmed they put up a plaque at a museum, and Elon said it will have fart thrusters via tweet. They have taken the corporate mentality of a plaque and a promise. They brought it into modern times by adding a follow up tweet by Elon. Yeah they don’t need a PR department. They get all the marketing they need from existing customers, and by giving away free roadsters fantasies

Also removing radar was a terrible move. It’s one thing if you never had it. The first time a Tesla runs over a kid without a radar, I would imagine the Defense for the victim would certainly ask is there anyway this tragedy could’ve been prevented if there was radar in the car. It’s one thing I have never had radar and be able to honestly answer we don’t know but we don’t think so. It’s another thing to have millions of miles of radar data showing that in fact it might’ve helped even in the slightest way to save the kids life. You would think if there was a supply chain issue tesla would’ve recalled early model teslas they did not purchase full self driving and Rob the radar is out of them. I still think it is a patent infringement issue where Tesla wants to be the only vision only Supplier and if they can make that work they would have a lock on it. It’s only my belief but I’m trying to find something that makes sense. But they did spend money on aero wheels, and cut weight by reducing insulation. I’ve got extended range but terrible road noise, which sadly sometimes my immersive sound can just barely overcome.

Elon Electrify Cuba
Umm, I don't think radar can see kids since they are not metal.
 

DocScott

Top-Contributor
TOO Supporting Member
Joined
Mar 6, 2019
Messages
619
Location
Westchester, NY
Country
Country
Tesla Owner
Model 3
Umm, I don't think radar can see kids since they are not metal.
Radar visibility has nothing to do with whether an object is metal.

Think about it: weather radars see rain!

EDIT: My statement was too strong. Radar visibility is affected by the composition of an object, and metals are particularly reflective. The same is true for visible light: metals are shiny, and you've got a better chance of catching a glint of sunlight off a distant metal object than a matte one! But it's certainly not the case that an object has to be metal in order for radar to see it.
 
Last edited:

Tesla4Me!

Member
Joined
Jul 29, 2017
Messages
23
Location
Germantown, MD
Country
Country
Tesla Owner
Model 3
Radar visibility has nothing to do with whether an object is metal.

Think about it: weather radars see rain!
Our car radar can see metal objects THROUGH rain. If it stopped at rain, it would not be able to see anything beyond the rain. I am pretty sure our car radar does not see people.
 

Madmolecule

Top-Contributor
Ludicrous Level Supporter
Joined
Oct 7, 2018
Messages
829
Location
Duluth, GA
Country
Country
Tesla Owner
Model 3
Radar can see metal objects THROUGH rain. And through people. It does not see people.
Radar can reflect off people. Unless they’re extremely dehydrated. Turns out they have a little bit of water in them. Radar is actually a great use for measuring water level. It has the ability to see through less dense things like foam and bring back the accurate water level. It actually sees the raindrops but it also sees past where there are no drops, which is tha majority. might help them make a windshield wiper that works. Actually LiDAR can also, it just doesn’t reflect to give you as accurate readings.

A major advantage of radar is it is not as susceptible to dirt and Pollen as a camera! on a vision only system I would hate for them to use the excuse they ran over the kid because the camera was dirty. What is the drivers responsibility and camera cleaning. Am I responsible to make sure it at least has 97% transmittance. How would I know that.
C511A9DF-D3CF-44CC-AF77-252CFE114232.jpeg
 
Last edited:

Tesla4Me!

Member
Joined
Jul 29, 2017
Messages
23
Location
Germantown, MD
Country
Country
Tesla Owner
Model 3
Our car radar can see metal objects THROUGH rain. If it stopped at rain, it would not be able to see anything beyond the rain. I am pretty sure our car radar does not see people.

Q. Can radar detect people, or will nearby people interfere with sensing the target?​

A higher frequency radar sensor (122 GHz) will more reliably detect people but is not intended to be used in personnel detection as a safety rated device. Depending on sensor sensitivity settings, a person could interfere with sensing the intended target if a person is near the sensor and within the sensor's direct field of view.
Tesla's radar is reported to be in the 76 -77 GHz range, so not as sensitive to water or people. A sample reference: https://www.bannerengineering.com/z.../radar-sensor-frequently-asked-questions.html
 

DocScott

Top-Contributor
TOO Supporting Member
Joined
Mar 6, 2019
Messages
619
Location
Westchester, NY
Country
Country
Tesla Owner
Model 3
Tesla's radar is reported to be in the 76 -77 GHz range, so not as sensitive to water or people. A sample reference: https://www.bannerengineering.com/z.../radar-sensor-frequently-asked-questions.html
I agree that Tesla radars and the AI interpreting them are not good at detecting pedestrians. But I think you're putting way too much emphasis on people not being metal. The radar reflectivity of things that are mainly water, like people, isn't quite as good as metal, but is still quite good. (In reply to your earlier comment: of course the radar is not stopped by rain. Even weather radar isn't stopped by rain, unless it's insanely heavy--if it were, all the radar could see would be the rain nearest to it, and if it were raining at the radar site, that would be pretty useless!)

There are two main reasons Tesla radar has trouble with pedestrians:

1) Pedestrians are much smaller than cars. That creates resolution issues.

2) Pedestrians move much more slowly than cars moving on a road.

#2 is the crucial one. Tesla radar is going to be getting returns from the environment all the time...ground level is a crowded place! Even the road surface itself may generate returns. So the radar needs a way of identifying what objects are important. It turns out with radar it's pretty easy to tell if an object is moving, via the Doppler shift. When radar bounces off a moving object, it comes back at a different frequency than it went out at. That's how police radar guns determine a vehicle's speed for instance (same with pitch speed in baseball). It's easy to note the different frequency, in the radar return, and thus know there's a moving object out there and how fast it's moving.

That means that Tesla radar is bad at detecting stationary objects, whether pedestrians, cars, trucks, or walls. It's programmed to ignore all those returns, because it's almost always getting returns like that. Most of the bad collisions Teslsas have had while on AP have been with stationary objects.

A pedestrian is nearly stationary, and thus Tesla radar/AI will generally ignore it, relying on vision instead.

In short, I agree with your main point: radar isn't being used in Teslas to detect pedestrians. But I disagree with your reason as to why; whether something is metal or not isn't the main issue. It's primarily how fast it's moving and to a lesser degree its size.
 

JasonF

Legendary Member
Joined
Oct 26, 2018
Messages
2,455
Location
Orlando FL
Country
Country
Tesla Owner
Model 3
Weather doppler radar returns reflectivity, velocity, range, and height measurements. Where that kind of applies here is they filter out ground interference by removing detected objects with very high reflectivity and zero velocity.

I don’t know how similar the radar Tesla uses is to that, but if it is, it sort of explains how Autopilot ends up hitting stationary objects in the road. It doesn’t see them because of how the filtering works.

Pedestrians would be hard to detect the same way pop up showers are hard to detect in weather - because they have very low velocity, brief reflectivity, and a very small cross section. So radar would see, then not see, then see again that pedestrian.
 
  • Agree
Reactions: DocScott

Tesla4Me!

Member
Joined
Jul 29, 2017
Messages
23
Location
Germantown, MD
Country
Country
Tesla Owner
Model 3
I agree that Tesla radars and the AI interpreting them are not good at detecting pedestrians. But I think you're putting way too much emphasis on people not being metal. The radar reflectivity of things that are mainly water, like people, isn't quite as good as metal, but is still quite good. (In reply to your earlier comment: of course the radar is not stopped by rain. Even weather radar isn't stopped by rain, unless it's insanely heavy--if it were, all the radar could see would be the rain nearest to it, and if it were raining at the radar site, that would be pretty useless!)

There are two main reasons Tesla radar has trouble with pedestrians:

1) Pedestrians are much smaller than cars. That creates resolution issues.

2) Pedestrians move much more slowly than cars moving on a road.

#2 is the crucial one. Tesla radar is going to be getting returns from the environment all the time...ground level is a crowded place! Even the road surface itself may generate returns. So the radar needs a way of identifying what objects are important. It turns out with radar it's pretty easy to tell if an object is moving, via the Doppler shift. When radar bounces off a moving object, it comes back at a different frequency than it went out at. That's how police radar guns determine a vehicle's speed for instance (same with pitch speed in baseball). It's easy to note the different frequency, in the radar return, and thus know there's a moving object out there and how fast it's moving.

That means that Tesla radar is bad at detecting stationary objects, whether pedestrians, cars, trucks, or walls. It's programmed to ignore all those returns, because it's almost always getting returns like that. Most of the bad collisions Teslsas have had while on AP have been with stationary objects.

A pedestrian is nearly stationary, and thus Tesla radar/AI will generally ignore it, relying on vision instead.

In short, I agree with your main point: radar isn't being used in Teslas to detect pedestrians. But I disagree with your reason as to why; whether something is metal or not isn't the main issue. It's primarily how fast it's moving and to a lesser degree its size.
I stand corrected on the details. Thanks for the detailed response.
 
  • Like
Reactions: DocScott

gaduser

Active member
Joined
Jun 5, 2019
Messages
30
Location
Novelty, Ohio
Country
Country
Tesla Owner
Model 3
You bring up a very good point here - there is ZERO data to show AP/FSD all by itself (since it doesn't exist) is safer than human-only drivers. Since there is no non-human-driver data to compare it with.

Reminds me of my favorite t-shirt logo:

"There are two kinds of people in the world: Those that can extrapolate from incomplete data"
That logo reminds me of another - "Beam me up Scotty, there's no sign of intelligent life down here".