Tesla replacing ultrasonic sensors with Tesla Vision

[Knightshade]

If I was to guess, a lot, I mean orders of magnitude, more EAP customers than P85D. P85D which was the most expensive of the 3 trims, with the largest delta ($25K) from the 85D (60 to 85D was $10K delta). Also, P85D was only sold for a fraction of time EAP was sold.

The EAP suit began March 2017, so the only eligible people at the time were specifically EAP buyers in the previous 6 months (because it had only been on sale 6 months at that point and only for new cars purchased in those 6 months with AP2 HW).

P85D was sold a lot longer than that.

It's certainly possible there were still more EAP buyers in those 6 months than in the couple years the P85D was sold (or vice versa), but I doubt it's a massive difference either way.

 

[whitex]

The EAP suit began March 2017, so the only eligible people at the time were specifically EAP buyers in the previous 6 months (because it had only been on sale 6 months at that point and only for new cars purchased in those 6 months with AP2 HW).

P85D was sold a lot longer than that.

It's certainly possible there were still more EAP buyers in those 6 months than in the couple years the P85D was sold (or vice versa), but I doubt it's a massive difference either way.

I don't think P85D was sold for a couple of years. October 2014 to Aug 2015 IIRC, P90D replaced it pretty quick, even before the refresh facia and AP2. Also, not sure which EAP suit you're talking about, since EAP was only offered starting late 2016, why would a lawsuit start in early 2017?

 

[Knightshade]

I don't think P85D was sold for a couple of years. October 2014 to Aug 2015 IIRC, P90D replaced it pretty quick, even before the refresh facia and AP2. Also, not sure which EAP suit you're talking about, since EAP was only offered starting late 2016, why would a lawsuit start in early 2017?

The lawsuit started 6 months after EAP went on sale because tesla had failed to deliver a bunch of features they said would already be there. They DID eventually deliver them (about a year after launch) but at just 6 months late the class action started up--- which Tesla later settled for a few hundred per owner (since they DID deliver, just a bit late).

FWIW I believe buyers of FSD from roughly March 2019 through mid-2020 at least (the period where they listed specific features with your purchase, and specifically said "coming later this year" for city streets driving) would be eligible for a similar lawsuit- and I'm honestly surprised we haven't had one yet... technically (in NA anyway) Tesla HAS delivered that now with fsdb, but again they gave a specific date for a while and were late on it.... this differs from the buyers of FSD before and after that period who were never promised any specific date for delivery of any specific feature.


As to the P85D, it was sold through early 2016 at least according to this which calls out the P85D as one of the just-discontinued-early-2016 models:

You can no longer buy Tesla's mid-range Model S

Tesla is trimming up its line-up ahead of the launch of its Model 3.

www.businessinsider.com

The RWD 85 models died sooner.

 

[whitex]

The lawsuit started 6 months after EAP went on sale because tesla had failed to deliver a bunch of features they said would already be there. They DID eventually deliver them (about a year after launch) but at just 6 months late the class action started up--- which Tesla later settled for a few hundred per owner (since they DID deliver, just a bit late).

I picked up a Model S in Dec 2016, which was already a few months in selling EAP. The AP computer was not even running for another 6-9 months after that (not even auto headlights which just required light level measurement), so I'm very sure nothing was delivered for EAP by March 2017 (though it was listed in the Design Studio as coming in Dec 2016 - see below).

As to the P85D, it was sold through early 2016 at least according to this which calls out the P85D as one of the just-discontinued-early-2016 models:

You can no longer buy Tesla's mid-range Model S

Tesla is trimming up its line-up ahead of the launch of its Model 3.

www.businessinsider.com

The RWD 85 models died sooner.

Here is a 2015 P90D:

2015 Tesla Model S P90D For Sale in Universal City, TX - 5YJSA1E42FF102885 - TrueCar

This 2015 Tesla Model S P90D is for Sale in Universal City, TX. Price: $37,850 - Mileage: 80,742 - Color: Gray- Transmission: Automatic - VIN: 5YJSA1E42FF102885

www.truecar.com

85's did not die sooner, they were sold along the side of 90's for a bit, as 90 was a $3K "range upgrade" for a bit, so you could order 85 or 90 cars (RWD or D), except for Performance which required the 90 upgrade and had an optional $10K Ludicrous mode.

 

[Knightshade]

I picked up a Model S in Dec 2016, which was already a few months in selling EAP.

EAP went on sale in October.

The lawsuit was announced in March 2017.

Thus 6 months from onsale to lawsuit....and as you point out only ~3 months from missing the promised delivery date to lawsuit

The AP computer was not even running for another 6-9 months after that (not even auto headlights which just required light level measurement), so I'm very sure nothing was delivered for EAP by March 2017 (though it was listed in the Design Studio as coming in Dec 2016 - see below).

That is the entire basis of the successful lawsuit so I'm not sure why you're repeating stuff I began the discussion pointing out?

Also why I mentioned those who bought FSD during the period Tesla had city streets "coming later this year" as the group having the best CURRENT chance of a succesful lawsuit- because it's the exact same principle- they were promised something by a specific date and it was not delivered to them.

Here is a 2015 P90D:

2015 Tesla Model S P90D For Sale in Universal City, TX - 5YJSA1E42FF102885 - TrueCar

This 2015 Tesla Model S P90D is for Sale in Universal City, TX. Price: $37,850 - Mileage: 80,742 - Color: Gray- Transmission: Automatic - VIN: 5YJSA1E42FF102885

www.truecar.com

Ok.

And here's a couple of 2016 P85Ds

Used 2016 Tesla Model S P85D AWD for Sale (with Photos) - CarGurus

Browse the best March 2024 deals on 2016 Tesla Model S P85D AWD vehicles for sale. Save $10,532 this March on a 2016 Tesla Model S P85D AWD on CarGurus.

www.cargurus.com

So when you claimed they were only sold "October 2014 to Aug 2015" that is not accurate.

They were sold into early 2016.



Point being, the EAP lawsuit began after only 6 months of EAP being on sale at all, and only ~3 months after missing the delivery date of EAP features.

The P85D was on sale from 2014 to early 2016.

So it's entire possible the # for a suit from each isn't terribly different.






as 90 was a $3K "range upgrade" for a bit, so you could order 85 or 90 cars (RWD or D), except for Performance which required the 90 upgrade and had an optional $10K Ludicrous mode.
[/QUOTE]

 

[whitex]

EAP went on sale in October.

The lawsuit was announced in March 2017.

Thus 6 months from onsale to lawsuit....and as you point out only ~3 months from missing the promised delivery date to lawsuit




That is the entire basis of the successful lawsuit so I'm not sure why you're repeating stuff I began the discussion pointing out?

Also why I mentioned those who bought FSD during the period Tesla had city streets "coming later this year" as the group having the best CURRENT chance of a succesful lawsuit- because it's the exact same principle- they were promised something by a specific date and it was not delivered to them.





Ok.

And here's a couple of 2016 P85Ds

Used 2016 Tesla Model S P85D AWD for Sale (with Photos) - CarGurus

Browse the best March 2024 deals on 2016 Tesla Model S P85D AWD vehicles for sale. Save $10,532 this March on a 2016 Tesla Model S P85D AWD on CarGurus.

www.cargurus.com

So when you claimed they were only sold "October 2014 to Aug 2015" that is not accurate.

They were sold into early 2016.



Point being, the EAP lawsuit began after only 6 months of EAP being on sale at all, and only ~3 months after missing the delivery date of EAP features.

The P85D was on sale from 2014 to early 2016.

So it's entire possible the # for a suit from each isn't terribly different.






as 90 was a $3K "range upgrade" for a bit, so you could order 85 or 90 cars (RWD or D), except for Performance which required the 90 upgrade and had an optional $10K Ludicrous mode.

[/QUOTE]
Btw, from Bing chatGPT

and

So it seems P85D was still available while P90D was out. So probably P90D was a P85D with a $3K range upgrade. In any case, even if you take the full Nov 2014-Feb 2016, it's not two years, not even a year an half.

 

[Knightshade]

I mean technically I said the couple years it was sold... 2014, 2015, and 2016-- just not all of the first or last ones

Point being (for the 3rd or 4th time now?) I doubt the # of people who purchased the P85D in that period is so MASSIVELY smaller than the # that bought EAP across just a few specific months at launch in 2016 that THAT is the reason one has a successful class action suit with a settlement and the other never even had a case filed on their behalf.

As FURTHER evidence of that- the P85D owners only in Norway DID get one filed on their behalf...and won a settlement.

Surely you'd admit the # of US P85D owners is larger than THAT?


Thus it's far more likely the P85D folks didn't have a case filed on their behalf in the US because they had a much weaker case under US law than the EAP owners did in their claim. Not because there just weren't enough of em.



(Oh, also-FYI ChatGPT is a terrible primary source, it's been repeatedly caught outright making things ups... even making up fake court cases as citations for claims it makes that aren't actually true, involving cases that never happened-- that it got it mostly right in this case is great (but then it's the same thing I'd already told you, WITH actual contemporary primary sources linked for you) but don't count on it--- lastly, the P85D was unveiled October 2014 same time EAP/HW2 was, though I don't know date of first delivery)

 

[joelliot]

I mean technically I said the couple years it was sold... 2014, 2015, and 2016-- just not all of the first or last ones

Point being (for the 3rd or 4th time now?) I doubt the # of people who purchased the P85D in that period is so MASSIVELY smaller than the # that bought EAP across just a few specific months at launch in 2016 that THAT is the reason one has a successful class action suit with a settlement and the other never even had a case filed on their behalf.

As FURTHER evidence of that- the P85D owners only in Norway DID get one filed on their behalf...and won a settlement.

Surely you'd admit the # of US P85D owners is larger than THAT?


Thus it's far more likely the P85D folks didn't have a case filed on their behalf in the US because they had a much weaker case under US law than the EAP owners did in their claim. Not because there just weren't enough of em.



(Oh, also-FYI ChatGPT is a terrible primary source, it's been repeatedly caught outright making things ups... even making up fake court cases as citations for claims it makes that aren't actually true, involving cases that never happened-- that it got it mostly right in this case is great (but then it's the same thing I'd already told you, WITH actual contemporary primary sources linked for you) but don't count on it--- lastly, the P85D was unveiled October 2014 same time EAP/HW2 was, though I don't know date of first delivery)

Of topic, but the bing chat actually sites sources that you can verify so it’s very different than ChatGPT.

 

[rgiaimo]

You need update 2023.6.9 or newer to get Vision park assist.

Turn on your rear view camera as you are pulling in and use the line that separates your driveway with your garage as a reference point for how far to pull in. You can even add a piece of tape to the ground as a reference point. This is what I do when I park in a spot where USS can't be used as a reference point. I mainly pull in enough so that I can open the trunk without it banging on the garage door (or vice versa, with either fully opened or closed).

Details on how to show rear view camera when going forward (should be similar for Model S):
Tesla replacing ultrasonic sensors with Tesla Vision

2023.6.9 does not have park assist for the Model S, no. It's still not there after the latest update last week. Rear camera when driving forward may be a solution.
Has anyone noticed that without 'creep', it's really difficult to keep the car moving at 1 MPH? Backing out of the driveway which has a slight incline is difficult without creep.

 

[JHCCAZ]

... anyone noticed that without 'creep', it's really difficult to keep the car moving at 1 MPH? Backing out of the driveway which has a slight incline is difficult without creep.

Just give it a little time, you'll get used to it and you'll probably end up preferring it. That's what happens with most new Tesla owners regarding one-pedal driving in general.

If you've ever driven a manual transmission car, it's hard then to complain about Tesla's system when creeping up an incline. It's just a little different, and you will become accustomed and comfortable with it shortly.

As an interesting thought experiment, just imagine if the whole world were accustomed to driving using the EV/Tesla system, and then someone introduced a car that moved forward (or back) by itself, unless you actively stop it. The internet would be filled with confusion and outrage. People would cry Mayhem and the CEO would be deemed a child-killer.

OTOH, I am very sympathetic with the problem of missing parking sensor functionality. Hopefully Tesla will get that solved soon.

 

[jeremymc7]

Turning off creep was the first thing I did with my first car. Not having it move is so much more functional. You just need a couple weeks to learn to appreciate it.

 

[stopcrazypp]

2023.6.9 does not have park assist for the Model S, no. It's still not there after the latest update last week. Rear camera when driving forward may be a solution.
Has anyone noticed that without 'creep', it's really difficult to keep the car moving at 1 MPH? Backing out of the driveway which has a slight incline is difficult without creep.

As others mentioned you'll get used to it as you drive with it a while.

One tip: pressing and holding (and releasing) brake pedal can put the car into and out of hold mode. You can use this to let the car slowly roll down, while having the brake partially applied. This is how I park when on hills.

 

[jeremymc7]

Besides doesn't one petal driving with hold maximize regen; regardless of how little? And reduce physical brake wear?

 

[Dewg]

A good video showing USS vs Vision. I was impressed with the backup camera, but the front camera needs work.

 

[whitex]

A good video showing USS vs Vision. I was impressed with the backup camera, but the front camera needs work.

On par for Elon's other brain farts. Reminds me of when Elon decided he can do BSM using USS, only to cleanse the Tesla website of any blind spot warnings for AP1 cars after they realized Elon never actually had anyone try USS for BSM before claiming the capability. Elon sold it with "we'll write the software later" attitude, only to fail miserably.

 

[Dan D.]

On par for Elon's other brain farts. Reminds me of when Elon decided he can do BSM using USS, only to cleanse the Tesla website of any blind spot warnings for AP1 cars after they realized Elon never actually had anyone try USS for BSM before claiming the capability. Elon sold it with "we'll write the software later" attitude, only to fail miserably.

I was reasonably impressed with the USS function in his video, and the backup camera worked with Tesla Vision to the extent he tested it.

The front camera is too unreliable and risky. But then it was a non-starter through software because it's all blind-spot. What made them think it wasn't?

 

[whitex]

I was reasonably impressed with the USS function in his video, and the backup camera worked with Tesla Vision to the extent he tested it.

The front camera is too unreliable and risky. But then it was a non-starter through software because it's all blind-spot. What made them think it wasn't?

I don't personally know anyone who works there anymore, but there is a good chance "they" didn't think it would work, only Elon did. Just like using parking sensors for BSM - Elon never stopped to think why they are called parking sensors (hint, they don't work well at high speeds with wind blowing over them at 70mph+). Elon could have asked some intern to verify, but no, he sells it first, then claims it was never promised.

 

[jeewee3000]

I was reasonably impressed with the USS function in his video, and the backup camera worked with Tesla Vision to the extent he tested it.

The front camera is too unreliable and risky. But then it was a non-starter through software because it's all blind-spot. What made them think it wasn't?

As a daily "non-USS-park-assist" user, I can confirm the front measurements are very unreliable.

If Tesla Vision (TV) was always too careful (by underestimating the distance) it could be relied on somewhat. However, sometimes it underestimates (by a lot) and sometimes it overestimates (by a lot).

Therefore I always have to carefully look at my surroundings to guesstimate if TV is "getting it" at that time. (getting it = seeing the object I'm seeing clearly)

Before I get jumped by posters telling me I should always carefully check my surroundings, of course I do. But the whole point of park assist is you can rely on it to do more tight manoeuvres where eyesight alone gives you insufficient data, such as "am I 5 inches away or 10 inches?" TV does not allow for this and therefore does not do what it is for. In other words: it is useless. (And no, this is not hyperbole. It is truly does not serve its purpose.)

Other gripes with TV park assist:

- when I park alongside a curb (from a sidewalk), I use my rear cameras to get the tires very close to the curb without hitting them (one inch max). This is very easy to do with the (excellent) camera placement. However, since TV park assist showed up it starts beeping like crazy telling me to "STOP" showing the curb as a solid red line right along the edge of my car. OK, great, it can see the curb, but by showing me the distance to this curb, even after I am parallel to it, it does NOT show me the distance to the vehicles in front and behind me. The distance reading just says STOP the whole time even as I approach the car in front (to get a litter closer). So again I have to guess the distance with human vision instead of Tesla Vision. Useless.

- the "park assist is loading" whenever I get in my car and put it in Drive, takes way too long, rendering TV useless. If I wait for TV to appear, bystanders throw me strange looks ("why is that guy not getting out of his parking spot?") I just start driving and I generally get a reading of TV after leaving my parking spot.

The thing is, TV park assist is supposed to be based on the occupancy network, and that same occupancy network is used as a basis for FSD.

If the occupancy network is that unreliable I cannot imagine FSD ever allowing hands off/eyes off autonomous driving.

Therefore I can only hope Tesla will improve the occupancy network accuracy/reliability over time with more and better AI training, so that the Tesla Vision features (park assist, FSD, etc) can reap the benefits.

Could somebody with a decent grasp on AI explain to me if further improvement is possible in this regard? Or should I not get my hopes up?

 

[Mullermn]

OK, great, it can see the curb, but by showing me the distance to this curb, even after I am parallel to it, it does NOT show me the distance to the vehicles in front and behind me. The distance reading just says STOP the whole time even as I approach the car in front (to get a litter closer).

Yeah, I think it would be really helpful if it was more specific about which distance it is currently showing you. I think some (though not all) of the ‘vision park assist is really inaccurate’ complaints are simply that the car is displaying a distance to a different object than the human assumes it is.

The thing is, TV park assist is supposed to be based on the occupancy network, and that same occupancy network is used as a basis for FSD.

If the occupancy network is that unreliable I cannot imagine FSD ever allowing hands off/eyes off autonomous driving.

I’m not sure this is a concern. You don’t need to understand the position of most entities on the road to particularly high precision to be able to drive a car - if you had to estimate the position of the car across the road to 1 inch precision how accurate would you be?

Park assist is trying to apply the same principle used in FSD to a much higher level. Of course if the accuracy doesn’t improve then smart summon etc aren’t coming back.

 

[Knightshade]

Yeah, I think it would be really helpful if it was more specific about which distance it is currently showing you. I think some (though not all) of the ‘vision park assist is really inaccurate’ complaints are simply that the car is displaying a distance to a different object than the human assumes it is.

I mean if it's only going to display ONE distance it should the to the nearest thing you're about to hit. Which does not seem to be the case consistently?