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

Tesla autopilot HW3

SVMike

Member
Aug 1, 2018
267
331
Silicon Valley
Not sure, I live vicariously through the reports of others.
From the autonomy presentation, I though Karpathy was talking about detecting non-signally lane changers, which implies they detect signalling lane changers.

I don't think the car sees other car's signals. Many times, when in AP, I see a car signal ahead of me as they are obviously going to move over into my lane. The car takes 2-4 seconds longer to start to slow down than an observant /courteous driver would take. It seems that the car ahead needs to be ~50% into my lane before my car starts to slow down.
 

dgatwood

Member
Dec 20, 2017
833
932
Sunnyvale, ca
Is the consensus, that unless you are on HW3 the that car not only will not visually show the cones but would also not react to the cones to go along with the 2019.36.1 software changes (presumably improved over whatever cone avoidance/navigation logic they had)?

FYI, in 2019.32.12.*, mine seems to sometimes detect and avoid (but not show) cones on HW2.5. However, it isn't nearly consistent enough to trust it.
 

J1mbo

Active Member
Aug 20, 2013
1,591
1,383
UK
Had a 3hr drive in really bad rain at the weekend. Overall, very impressed with the car (M3/FSD/32.12.2) - to the point that I felt more comfortable driving with AP on than off. There were a few moments where it handed back to me - mostly when a vehicle cut in front and threw up a lot of spray - and functionality was restricted for a lot of of the trip, in particular blindspot detection.

Checking out the TeslaCam video shows exactly why HW3 will only be L4/5 in good conditions.

Left Cam - seriously affected by vibration and light blooming, no wonder blindspot / lane change was disabled


Main Cam is amazingly clear in comparision, but check out all those hot pixels on the "automotive grade" cam :(


When AP finally gave up, behaviour was predictable, in that AP suddenly deactivated due to lack of vision. But, AP (main cam) had better vision than I did, so all I could do was slow down and follow the tail lights of the car ahead until visibility improved, at which point AP quickly became available again. Am thinking that warning, slowing down and following tail lights for 5-10 seconds would be a better behaviour from AP (given the extra safety of the radar) whilst handing back a bit more gracefully.
 

Saghost

Well-Known Member
Oct 9, 2013
8,217
7,010
Delaware
Had a 3hr drive in really bad rain at the weekend. Overall, very impressed with the car (M3/FSD/32.12.2) - to the point that I felt more comfortable driving with AP on than off. There were a few moments where it handed back to me - mostly when a vehicle cut in front and threw up a lot of spray - and functionality was restricted for a lot of of the trip, in particular blindspot detection.

Checking out the TeslaCam video shows exactly why HW3 will only be L4/5 in good conditions.

Left Cam - seriously affected by vibration and light blooming, no wonder blindspot / lane change was disabled


Main Cam is amazingly clear in comparision, but check out all those hot pixels on the "automotive grade" cam :(


When AP finally gave up, behaviour was predictable, in that AP suddenly deactivated due to lack of vision. But, AP (main cam) had better vision than I did, so all I could do was slow down and follow the tail lights of the car ahead until visibility improved, at which point AP quickly became available again. Am thinking that warning, slowing down and following tail lights for 5-10 seconds would be a better behaviour from AP (given the extra safety of the radar) whilst handing back a bit more gracefully.

Warning, following for several seconds using the best available data, and gracefully handing over control would be a level 3 solution to the situation.

Popping the flashers and pulling into the shoulder and stopping as the visibility dropped below the manageable level would be a level 4 solution.
 

J1mbo

Active Member
Aug 20, 2013
1,591
1,383
UK
Popping the flashers and pulling into the shoulder and stopping as the visibility dropped below the manageable level would be a level 4 solution.

Looking at the left cam, I very much doubt that would be possible to do safely with the current HW suite. Neither vision or ultrasonic would be trustworthy enough to get you there (imo).
 

S85D

Member
Jun 2, 2015
624
663
Austin, TX
While I am also of the opinion that AP2 will get HW3, we must note Elon also promised MCU2 retrofits to MCU1 cars (never happened) as well as Early Access Program invitations to early FSD buyers (never happened). Also there is a very recent service visit where the service advisor was adamant AP2 will not get HW3 even when pressed. Going even further back Tesla promised lighted vanity mirror retrofits to all Model S (never happened) to keep people taking delivery without.

As said, I do think AP2 will get HW3, but it isn’t completely out there as far as history goes to suggest some risk. Another question is how long will it take. One data point is the P100D retrofit that was sold in Q3/2016 to keep people taking delivery of their P90Ds. It took over a year to happen after P100D was released.

Don't forget that after Early Adopters did not get the invitation to join the EAP program, Tesla promised that we would be the first to get new firmware releases. Yet another promise broken.

You can always tell when Elon is lying, his mouth is moving...
 

electronblue

Active Member
Oct 1, 2018
2,325
2,411
Earth
Don't forget that after Early Adopters did not get the invitation to join the EAP program, Tesla promised that we would be the first to get new firmware releases. Yet another promise broken.

You are right, that was another broken promise. Smart Summon went to all FSD owners at the same, no priority to early adopters who indeed were supposed to get priority as compensation for having paid more than late adopters.

And now that HW3 is getting priority, there is no more possibility to give early FSD owners any priority anyway. Not that they are even trying: no AP2 retrofits yet...

So many lies.
 

timewasted

Member
Jun 29, 2017
196
530
Las Vegas, NV
Smart summon was absolutely an end of quarter deferred revenue recognition release. Expect to see another half baked feature released around the last week of December as well so that the “look at how profitable we are!” story can keep going.

As an aside, I put in a request for my AP3 retrofit (from AP2.0) via the app. I’m not going to bother updating the thread when they cancel my appointment because most rational people know that’s what’ll happen, but I’ll at least give them an opportunity to do something right and post if they manage to come through on it.
 

jebinc

M3 LR AWD w/FSD and white premium interior
Jun 19, 2019
3,412
1,688
Seattle area
Smart summon was absolutely an end of quarter deferred revenue recognition release. Expect to see another half baked feature released around the last week of December as well so that the “look at how profitable we are!” story can keep going.

As an aside, I put in a request for my AP3 retrofit (from AP2.0) via the app. I’m not going to bother updating the thread when they cancel my appointment because most rational people know that’s what’ll happen, but I’ll at least give them an opportunity to do something right and post if they manage to come through on it.
I believe the "stimulus" for 4Q sales may have been releasing 2020 cars in 2019 (pulling early 1Q20 orders into 4Q19). In the past, the Tesla year change occurred at the start of each year, and not in the Fall as other automakers do.
 

mongo

Well-Known Member
May 3, 2017
13,034
39,245
Michigan
Smart summon was absolutely an end of quarter deferred revenue recognition release. Expect to see another half baked feature released around the last week of December as well so that the “look at how profitable we are!” story can keep going.

Smart summon only contributed $30 million of the $150 million Q3 profit. Was not needed for the profitable story to continue.
$30 Million: Tesla, Inc. (TSLA) Q3 2019 Earnings Call Transcript | The Motley Fool
$150 Million: https://ir.tesla.com/static-files/47313d21-3cac-4f69-9497-d161bce15da4

Hope your AP3 install happens!
 

electronblue

Active Member
Oct 1, 2018
2,325
2,411
Earth
Smart summon only contributed $30 million of the $150 million Q3 profit. Was not needed for the profitable story to continue.
$30 Million: Tesla, Inc. (TSLA) Q3 2019 Earnings Call Transcript | The Motley Fool
$150 Million: https://ir.tesla.com/static-files/47313d21-3cac-4f69-9497-d161bce15da4

Let’s face it: Rushing a wide firmware release at the end of the quarter likely means one thing with Tesla. It may not have been necessary in the end, but they may not have known that for sure — and also they probably wanted to show as big a profit as possible.

@timewasted is in my view right to expect another revenue recognition firmware push at the end of a future quarter. Whether or not they are far along enough to push it at the end of Q4/2019 remains to be seen. Smart Summon was of course delayed many times.
 
  • Disagree
Reactions: Dave EV

mongo

Well-Known Member
May 3, 2017
13,034
39,245
Michigan
Let’s face it: Rushing a wide firmware release at the end of the quarter likely means one thing with Tesla. It may not have been necessary in the end, but they may not have known that for sure — and also they probably wanted to show as big a profit as possible.

@timewasted is in my view right to expect another revenue recognition firmware push at the end of a future quarter. Whether or not they are far along enough to push it at the end of Q4/2019 remains to be seen. Smart Summon was of course delayed many times.

I don't know that I would call it rushed, given the delays it had already experienced. Tesla seems to favor release of the min viable product to fail fast and improve from there. While not {insert superlative here}, it may have hit the limit of improvement for purely internal development and needed to be deployed widely to collect more data.

At some point, you need to deploy.
 
  • Disagree
Reactions: electronblue

Saghost

Well-Known Member
Oct 9, 2013
8,217
7,010
Delaware
I don't know that I would call it rushed, given the delays it had already experienced. Tesla seems to favor release of the min viable product to fail fast and improve from there. While not {insert superlative here}, it may have hit the limit of improvement for purely internal development and needed to be deployed widely to collect more data.

At some point, you need to deploy.


I think it was clearly a case where they needed more data to improve it - that’s why v10 went the hw3 cars first, and why they went through several updates over just a few days with visible improvements in performance from version to version IMHO.
 

electronblue

Active Member
Oct 1, 2018
2,325
2,411
Earth
I don't know that I would call it rushed, given the delays it had already experienced. Tesla seems to favor release of the min viable product to fail fast and improve from there. While not {insert superlative here}, it may have hit the limit of improvement for purely internal development and needed to be deployed widely to collect more data.

At some point, you need to deploy.

Had this been a usual case, it would have taken a couple of weeks for all cars to get V10, as they went through the usual drip.

Instead everyone got it on the last days of the quarter — after which Tesla immediately recognized 30 million.

There was no HW3 preference either, in reality. Literally everyone got it within a couple of days at the end of the quarter.

Don’t be naive. This is Tesla and the end of the quarter. It isn’t like there isn’t history.
 

electronblue

Active Member
Oct 1, 2018
2,325
2,411
Earth
I think it was clearly a case where they needed more data to improve it - that’s why v10 went the hw3 cars first, and why they went through several updates over just a few days with visible improvements in performance from version to version IMHO.

An alternative view: They went through several rapid updates because they had rushed V10 to the entire fleet at the end of the quarter — and needed to patch things up afterwards.

The usual way would have seen the update take several weeks over the entire fleet.
 

Saghost

Well-Known Member
Oct 9, 2013
8,217
7,010
Delaware
An alternative view: They went through several rapid updates because they had rushed V10 to the entire fleet at the end of the quarter — and needed to patch things up afterwards.

The usual way would have seen the update take several weeks over the entire fleet.

I’d be more inclined towards that view if we hadn’t seen the improvement in smart summon as the revisions rolled.
 

mongo

Well-Known Member
May 3, 2017
13,034
39,245
Michigan
Had this been a usual case, it would have taken a couple of weeks for all cars to get V10, as they went through the usual drip.

Instead everyone got it on the last days of the quarter — after which Tesla immediately recognized 30 million.

There was no HW3 preference either, in reality. Literally everyone got it within a couple of days at the end of the quarter.

Don’t be naive. This is Tesla and the end of the quarter. It isn’t like there isn’t history.

End of quarter is just as valid a release target as any other time.
Why would there be a HW3 preference (other than the FSD get features early bonus)?
No need for the staged general deployment of AP features. This is not a safety critical feature. It is a dead man switch low speed limited use feature. Is your claim that the October improvements were due to an extra week or two on internal development instead of the 500,000+ uses of the initial release?

To original point:
Did it help them to claim deferred revenue by having it out to the whole feet? Yes
Was that revenue critical to the quarterly profit? No
 

electronblue

Active Member
Oct 1, 2018
2,325
2,411
Earth
@mongo Tesla stages most of their deployments. (HW3 priority was mentioned by someone else, I noted there was none.) This time they did not. Just a coincidence? I think not.

As I said, Tesla may not have known the final numbers anyway when they decided to launch — it all adds up to the rush of the final days of the quarter and it seems likely they wanted to make sure.

Mind you I’m not saying launching Smart Summon at the end of the quarter would be only about recognized revenue. It potentially has other benefits from a stock market perspective as well. :)
 

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
Top