Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Tesla autopilot HW3

This site may earn commission on affiliate links.
Other than not steer...
Sure, you could ascribe a level of inattentiveness with more data, but does that matter? Had hands on wheel and crashed car, had hands off wheel and crashed car, what's the difference?

It matters a lot because a possible moment’s inattentiveness can be made sound like one lasting for much longer — an attentive driver that could have had a momentary lapse (or blinded by the sun) made sound like they were not attentive during the whole drive.

I’m not so worried about the immediate seconds leading up to the crash being misreperesented because as you say there the driver’s inattentiveness seems likely if they really took no action. But this lack of detection can and in my view has been used in these instances to form a narrative over a period of time... and that is much more likely to be completely misleading when not put into context properly.

With such narrative, every single drive most of us (without defeat devices) have on AP where we hold the steering wheel with both hands as recommended can be made sound like we didn’t hold the wheel repeatedly and thus were dangerous drivers, when in fact the only issue was that the torque sensing didn’t detect our two hands. Ironically those dangerously only using one hand or a defeat device would sound like better drivers in this narrative.

Tesla could explain that the hand detection is far from perfect but of course they never do in these instances. One suspected reason would be they feel it helps the narrative to cast more suspicion on the driver instead. Hands were repeatedly not detected for many, many miles...
 
Last edited:
Tesla does not say that hands were not on the wheel, they say hands were not detected...

Except that isn't what they wrote;)

It does mean something, though Tesla is too polite to be blunt: that the driver took no action before the crash to avoid it, neither braking nor swerving. They didn't pay attention to driving (or were incapacitated) and crashed into the stationary object.

How can hands be detected if there is no detection device for hands on the steering wheel? All Tesla knows is the last time torque was applied to the steering wheel.

To say "hands not detected" implies that there is an accurate means to detect hands at the steering wheel.

To give Tesla some credit they have made some important changes. The nag now says to apply a slightly turning motion, and doesn't say hold the steering wheel anymore.

Tesla also didn't immediately blog about the latest fatality accident. Instead it looks like they're playing nice with the investigators, and they're not jumping to defend themselves.

As to why the driver didn't take action in the Mountain View accident. I think it's hard for those that weren't there to comprehend how quick the entire thing may have happened.

Here is a similar situation where luckily the driver was ready.

It's BACK! After 6 months of working fine, 2019.5.15 drives at barriers again : teslamotors

That's awfully fast so I could see how a second or two of inattention may have cost him his life.
 
Tl:dr This the HW3 thread, I hear where you are coming from, and can discuss elsewhere, but I think it is (another) agree to disagree.

Okay, so some are talking specific accident(s), that helps.
IMO, Tesla provides the data they have, how people interpret it is outside their control.
If the data showed an attempt to avoid, or the car suddenly veering, Tesla would report that.
Nothing is gained by classifying them as an general attentive driver, or otherwise.
An accident only takes a moment, any moment. One can be inattentive and lucky, or attentive, sneeze, and be unlucky.

(sort of OT)
If suprcruise had (has) data logging, is it 100% to say 'driver was not looking at the road vs 'driver was not determined to be looking at the road?'. And if so, does that indicate their hands were available to take action? (Burger and drink driving).
 
  • Helpful
Reactions: scottf200
One of my hot buttons is people who single out Tesla for industry wide issues and accuse Tesla of somehow not caring because they haven't solved a problem that exists for most modern cars.

Consider the tractor trailer crossing the road. This is a trucking standards problem. If side impact skirts were required there would be 3 beneficial effects:

1) Less fuel would be burned because they reduce aerodynamic drag.
2) Fewer people would die in crashes. Airbags, seat belts, and crumple zones would become effective again and the decapitation crash eliminated.
3) There would be fewer crashes. Metal side skirts would set off automatic emergency braking systems which, currently, often miss a trailer due to seeing the road below.

If you are at high speed and the car in front of you changes lanes to uncover a stopped vehicle, today's car radar systems (not just Tesla) may not see the stopped vehicle. If a vehicle is moving and slows to a stop, radar can track that and distinguish it from the road. When a stopped vehicle is uncovered, the system is presented with a problem. Is this a stopped vehicle or an uphill road segment? The Doppler shifts are the same. If the car is set to brake then there will be a lot of dangerous phantom braking due to rising roadways or to low signs or signs on the side of the road where the road turns so the signs are in the path. There is hope for this problem and it is HW3 related. It involves image recognition of the stopped vehicle activating emergency braking. Right now I don't know of any car that does this. I would expect Tesla to be one of the first but this is tricky since phantom braking is also dangerous.
 
One of my hot buttons is people who single out Tesla for industry wide issues and accuse Tesla of somehow not caring because they haven't solved a problem that exists for most modern cars.

Consider the tractor trailer crossing the road. This is a trucking standards problem. If side impact skirts were required there would be 3 beneficial effects:

1) Less fuel would be burned because they reduce aerodynamic drag.
2) Fewer people would die in crashes. Airbags, seat belts, and crumple zones would become effective again and the decapitation crash eliminated.
3) There would be fewer crashes. Metal side skirts would set off automatic emergency braking systems which, currently, often miss a trailer due to seeing the road below.

If you are at high speed and the car in front of you changes lanes to uncover a stopped vehicle, today's car radar systems (not just Tesla) may not see the stopped vehicle. If a vehicle is moving and slows to a stop, radar can track that and distinguish it from the road. When a stopped vehicle is uncovered, the system is presented with a problem. Is this a stopped vehicle or an uphill road segment? The Doppler shifts are the same. If the car is set to brake then there will be a lot of dangerous phantom braking due to rising roadways or to low signs or signs on the side of the road where the road turns so the signs are in the path. There is hope for this problem and it is HW3 related. It involves image recognition of the stopped vehicle activating emergency braking. Right now I don't know of any car that does this. I would expect Tesla to be one of the first but this is tricky since phantom braking is also dangerous.

I think there's one. Subaru uses the unique Eyesight stereo camera based system for adaptive cruise and the rest of the driver assistance and safety systems - without radar I believe.

The tests I've seen show it having a better record than most anything else with stopped cars and obstacles on the road.

But in general I agree with what you said. :)
 
I think there's one. Subaru uses the unique Eyesight stereo camera based system for adaptive cruise and the rest of the driver assistance and safety systems - without radar I believe.

The tests I've seen show it having a better record than most anything else with stopped cars and obstacles on the road.

But in general I agree with what you said. :)
That's very encouraging to hear.
 
Consider the tractor trailer crossing the road. This is a trucking standards problem. If side impact skirts were required there would be 3 beneficial effects:

1) Less fuel would be burned because they reduce aerodynamic drag.
2) Fewer people would die in crashes. Airbags, seat belts, and crumple zones would become effective again and the decapitation crash eliminated.
3) There would be fewer crashes. Metal side skirts would set off automatic emergency braking systems which, currently, often miss a trailer due to seeing the road below.
Those were my thoughts as well, until I came across reasoning from some of the truckers. The difference is that trailers in America are significantly longer than in Europe (so is the space between axles) and any permanent installation on the side would have a potential of them being stuck on some of the railway crossings. You see a lot of "soft" skirts and curtains already being retrofitted on the trailers, but these are just to provide better aerodynamics, not to prevent deadly accidents from happening.
 
Those were my thoughts as well, until I came across reasoning from some of the truckers. The difference is that trailers in America are significantly longer than in Europe (so is the space between axles) and any permanent installation on the side would have a potential of them being stuck on some of the railway crossings. You see a lot of "soft" skirts and curtains already being retrofitted on the trailers, but these are just to provide better aerodynamics, not to prevent deadly accidents from happening.
Sounds like a very solvable engineering problem to me in return for saving lives. Perhaps I am wrong but I can imagine one that easily is pushed up with a lower half pushing up inside an outer section.

A friend is the lawyer who caused trucks to have reflective stripes down the side. For years the trucking industry fought it as too expensive. Then a truck was stalled across a road at night. Several girls were decapitated. My friend was the lawyer for the families. Now trucks have stripes to help visibility at night. I expect the trucking industry will fight side impact skirts. I also expect that, like seat belts, we will look back one day and wonder what took so long.
 
Here’s what we need someone to do:

1) Take out an $80k+ loan.
2) Use said loan to purchase AP3 equipped vehicle.
3) Disassemble enough of the vehicle to get lots of quality pictures.
3a) Optionally fly verygreen/bigdog/wk057/etc to you to assist with the process as well as any “extra” work.
4) Put car back together.
5) Really put Elon’s claim that you can return any car for any reason within the first 7 days for a full refund.

Who here will volunteer for tribute?!

PS: I call “not it”
 
start with the option code decoders. if it shows aph4, remove the glove box to peek at the unit.
I logged into My Tesla account and using web inspector searched for APH3 and APH4, only hit's were this:
upload_2019-3-28_15-58-56.png
 
  • Informative
Reactions: daktari
So Elon just confirmed anyone with FSD will get Hardware 3. I guess this puts those rumors to rest.

Anyone who purchased full self-driving will get FSD computer upgrade for free. This is the only change between Autopilot HW2.5 & HW3. Going forward “HW3” will just be called FSD Computer, which is accurate. No change to vehicle sensors or wire harness needed. This is v important. sc on Twitter
 
  • Like
Reactions: Kant.Ing