Firmware Build v9.0 2019.24.4 73fb1ab (7/19/2019)

  • SUPPORT THE SITE AND ENJOY A PREMIUM EXPERIENCE!
    Welcome to Tesla Owners Online, four years young! For a low subscription fee, you will receive access to an ad-free version of TOO. We now offer yearly memberships! You can subscribe via this direct link:
    https://teslaownersonline.com/account/upgrades

    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!

  • It's OK to discuss software issues here but please report bugs to Tesla directly at servicehelpna@teslamotors.com if you want things fixed.

Did you install 2019.24.4?


  • Total voters
    373

Mike

Legendary Member
Joined
Apr 4, 2016
Messages
2,449
Location
Batawa Ontario
Country
Country
Since running this version, the handoff from Spotify on my phone to streaming it thru the cars stereo system is no longer reliable.

I have had to force the car to recognize the phone (no issues with phone as key FOB) a number of times.

Or, the music will play for literally half a second and then go silent, with the UI showing that there is no source available.

Detailed message to Tesla has been sent.
 
Joined
Feb 23, 2019
Messages
15
Location
Boston
Tesla Owner
Model 3
Country
Country
I use NoA heavily on my commute, but almost always take over at interchanges since it would go slow enough to annoy me and the person behind me. The steering was very herky-jerky through cloverleaf ramps as well. Yesterday on my commute, I was paying attention to the road in front of me, but was a little spaced out and didn't think to take over for the I-90 to I-95 interchange. I was pleasantly surprised when it handled it smoothly and at a reasonable speed much closer to how I'd drive.

This is a sample size of one, but I'll definitely be trying again. Anyone else notice an improvement here?
 

evannole

Active member
Joined
Jun 17, 2018
Messages
138
Location
Marietta, GA
Tesla Owner
Model 3
Country
Country
Alas, I spoke too soon with regard to phantom braking on 24.4. This morning, it was back - not hugely so but enough to be annoying.

Here's hoping that Elon's reference to "improved Autopilot on the highway" includes the virtual elimination of this problem. I really could not care less about new features at this point. I just want this one to work properly. My wife's 2016 Volvo XC60 has only cameras and radar, and while its TACC is not as good as Tesla's when Tesla's is working properly, it also has never once exhibited phantom braking.

I am *this close* to springing for the HW3 upgrade for $3,000 while I still can. While I don't really care about FSD, I do think that HW3 will probably make EAP perform better and might eliminate phantom braking completely. On the other hand, I am reluctant to blindly part with my money when this problem keeps happening - and, frankly, if HW3 is required to get EAP working properly and as advertised, with no phantom braking, then the upgrade should be provided free of charge.
 

msjulie

Top-Contributor
TOO Supporting Member
Joined
Feb 6, 2018
Messages
456
Location
San Fran Bay Area, Ca
Tesla Owner
Model 3
Country
Country
Completely agree. At some point in time the reliability of the system takes priority over adding anything new. If you can't solve the reliability problem, the value of the system becomes zero.
I continue to wonder if for 'controlled access roads'* that adding in peer-to-peer communication between vehicles is needed to really be trusted and safe. Only cooperating vehicles can be on some roads..

*what I mean by controlled access is any road or lanes? that have been designated for smart communicating vehicles. Can't trust other people, bicycles or pedestrians, horses, etc really and will any car ever really be safe enough to drive where non-smart transit is going?
 

MelindaV

☰ > 3
Moderator
Joined
Apr 2, 2016
Messages
10,008
Location
Vancouver, WA
Tesla Owner
Model 3
Country
Country
I continue to wonder if for 'controlled access roads'* that adding in peer-to-peer communication between vehicles is needed to really be trusted and safe. Only cooperating vehicles can be on some roads..

*what I mean by controlled access is any road or lanes? that have been designated for smart communicating vehicles. Can't trust other people, bicycles or pedestrians, horses, etc really and will any car ever really be safe enough to drive where non-smart transit is going?
and how do you keep the non-smart communicating vehicles out of that road/lane? on my commute, there is a 3 mile stretch of HOV that is to be restricted for 3 hours a day (3p-6p) monday-friday. It is the only HOV section of freeway in all of Oregon, and i sit in the adjacent lane each afternoon watching 3/4 of the cars in that lane being single people.
 

msjulie

Top-Contributor
TOO Supporting Member
Joined
Feb 6, 2018
Messages
456
Location
San Fran Bay Area, Ca
Tesla Owner
Model 3
Country
Country
and how do you keep the non-smart communicating vehicles out of that road/lane? on my commute, there is a 3 mile stretch of HOV that is to be restricted for 3 hours a day (3p-6p) monday-friday. It is the only HOV section of freeway in all of Oregon, and i sit in the adjacent lane each afternoon watching 3/4 of the cars in that lane being single people.
Same here in Bay Area, lots of car pool scofflaws

Re controlled access: I don't imagine it would be easy to get any of this done as the lanes can't be easy enough to drive into. My comment was more about my faith in smart cars and less-smart humans actually successfully on the same roads happening as soon as we might wish
 

JeanDeBarraux

Active member
Joined
Feb 18, 2019
Messages
62
Location
Barraux A-R-A
Tesla Owner
Model 3
Country
Country
Alas, I spoke too soon with regard to phantom braking on 24.4. This morning, it was back - not hugely so but enough to be annoying.

Here's hoping that Elon's reference to "improved Autopilot on the highway" includes the virtual elimination of this problem. I really could not care less about new features at this point. I just want this one to work properly. My wife's 2016 Volvo XC60 has only cameras and radar, and while its TACC is not as good as Tesla's when Tesla's is working properly, it also has never once exhibited phantom braking.

I am *this close* to springing for the HW3 upgrade for $3,000 while I still can. While I don't really care about FSD, I do think that HW3 will probably make EAP perform better and might eliminate phantom braking completely. On the other hand, I am reluctant to blindly part with my money when this problem keeps happening - and, frankly, if HW3 is required to get EAP working properly and as advertised, with no phantom braking, then the upgrade should be provided free of charge.
I'm not sure if I'm on HW2.5 or HW3 (I read somewhere all European M3s are on HW3, but this needs to be confirmed), but since I upgraded to 2019.24.4, I've started seeing the "phantom braking" others have reported. Every time this happened was after I had just passed a bicycle. This hadn't happened before.

Has anyone relied on AutoPilot to take their car through the narrow lanes in the toll plaza at one of the bridges in the SF Bay area, or similar?
I actually have let the car drive through the toll plaza once or twice since the upgrade and it worked fairly well, even though the lane marking are completely absent. Both times, the car stuck to the left lane and followed the car in front through the toll plaza.
 

tivoboy

Top-Contributor
Joined
Mar 24, 2017
Messages
608
Location
Palo Alto, CA
Country
Country
Completely agree. At some point in time the reliability of the system takes priority over adding anything new. If you can't solve the reliability problem, the value of the system becomes zero.

Internally I wonder if this is Musk's arrogant insistence that vision systems alone can "solve" the self driving problem without reliance on other types of imaging like LIDAR. Unless Tesla can fix their phantom brake issue, the value of TACC, autosteer, and the future of FSD are all moot. It's still a great car but it's no longer a technology play and Tesla's lead over the traditional ICE competitors becomes manufacturing (SOC), electronic components and a willingness to break the traditional car mold rather than automated driving.
It does bring up a point I've made here before. Why not just get SOME of the things right and REALLY RIGHT first, then move on to the much more lofty aspirations like FSD? I have the feeling they have the huge aspirational goal out there, since they've told customers they can do it, since they've told THE STREET they will do it by X, since they've built a lot of their value and appeal on at some date, X, being able to have the car go from A (insert any point in the USA) to B (insert a different point in the USA) all on it's own. Forget the regulatory implications - which I've said before are going to take much longer than current projections, but it's the MOST complicated solve to do.

Why not just NOA (or at a minimum TACC with lane changes) REALLY rockin'!. If I as a driver could be confident that 99% of the time the car on the highway, could do 100% of the driving, up to and including gettting over close to my exit, that would be MUCH better than what we have now. I'm just not satisfied or frankly confident in the cars ability AND CONFIDENCE to make appropriate lanes changes, take approprirate initiative, accelerate when it's needed, de-celerate when that is a better solve for the objective or current situation and just do the damb driving.

I have the feeling that the above answer/solve is being lost in the objective to solve it all (FSD) and it's having a negative impact on just getting the basics right.
 

tivoboy

Top-Contributor
Joined
Mar 24, 2017
Messages
608
Location
Palo Alto, CA
Country
Country
Has anyone relied on AutoPilot to take their car through the narrow lanes in the toll plaza at one of the bridges in the SF Bay area, or similar?
Uh, ZERO chance I would try that.. there are concrete medians there that would create some serious damage to a rim, easily and often. I'd also be VERY concerned that the car might mis-interpret the sudden prescence of the toll both, on either side and make a last second sudden adjustment - into the OTHER side of the toll booth.

I wonder if Tesla is even testing this out as a feature to offer?
 

Mr. Spacely

Well-known member
TOO Supporting Member
Joined
Feb 28, 2019
Messages
365
Location
St. Petersburg, FL
Tesla Owner
Model 3
Country
Country
Has anyone relied on AutoPilot to take their car through the narrow lanes in the toll plaza at one of the bridges in the SF Bay area, or similar?
I live in Florida and have taken my car through narrow tool plazas on AutoPilot with "SunPass" toll reader at 30-40 MPH hundreds of times without incident. The car will disengage out of NOA and go to regular Auto Pilot through the booth and then reengage. Pretty slick...
 
Joined
May 8, 2018
Messages
20
Location
Redmond wa
Tesla Owner
Model 3
Country
Country
I continue to wonder if for 'controlled access roads'* that adding in peer-to-peer communication between vehicles is needed to really be trusted and safe. Only cooperating vehicles can be on some roads..

*what I mean by controlled access is any road or lanes? that have been designated for smart communicating vehicles. Can't trust other people, bicycles or pedestrians, horses, etc really and will any car ever really be safe enough to drive where non-smart transit is going?
I think about this a lot as I drive around. I look at driving scenarios that humans handle with relative ease and question if any automated systems can ever really infer what is going on and ever be able to effectively duplicate the human driver.

A perfect example is a car making a left hand turn about 100 yards in front of you. I can see the gaps in traffic, I can infer when the driver will make the left hand turn and I can be prepared should they decide not to go through a gap in traffic and decide not to make the turn and vacate my lane. All the computer sees is a car stopped 100 yards in front of me. I will maintain speed up to a point where I know/don't know if they are going to take the turn, if they do vacate the lane through an anticipated traffic gap, I can continue at speed and smoothly. The computer just brakes, then accelerates based on an obstacle or not. No inference or contextual knowledge applied and a much rougher trip.
 
Joined
Feb 10, 2019
Messages
10
Location
Tempe, AZ
Tesla Owner
Model 3
Country
Country
A perfect example is a car making a left hand turn about 100 yards in front of you. I can see the gaps in traffic, I can infer when the driver will make the left hand turn and I can be prepared should they decide not to go through a gap in traffic and decide not to make the turn and vacate my lane. All the computer sees is a car stopped 100 yards in front of me. I will maintain speed up to a point where I know/don't know if they are going to take the turn, if they do vacate the lane through an anticipated traffic gap, I can continue at speed and smoothly. The computer just brakes, then accelerates based on an obstacle or not. No inference or contextual knowledge applied and a much rougher trip.
This is exactly a what I have seen and thought about too. I see that car turning in front of me and I can tell immediately that it will clear my path well before I get there, but currently autopilot cannot tell that. It just sees an object in my lane and slams on the brakes.

Hopefully with further training of neural network and maybe needing HW3, this situation can also be correctly interpreted.
 

Ze1000

Active member
Joined
May 21, 2018
Messages
130
Location
South Florida
Tesla Owner
Model 3
Country
Country
It does bring up a point I've made here before. Why not just get SOME of the things right and REALLY RIGHT first, then move on to the much more lofty aspirations like FSD? I have the feeling they have the huge aspirational goal out there, since they've told customers they can do it, since they've told THE STREET they will do it by X, since they've built a lot of their value and appeal on at some date, X, being able to have the car go from A (insert any point in the USA) to B (insert a different point in the USA) all on it's own. Forget the regulatory implications - which I've said before are going to take much longer than current projections, but it's the MOST complicated solve to do.

Why not just NOA (or at a minimum TACC with lane changes) REALLY rockin'!. If I as a driver could be confident that 99% of the time the car on the highway, could do 100% of the driving, up to and including gettting over close to my exit, that would be MUCH better than what we have now. I'm just not satisfied or frankly confident in the cars ability AND CONFIDENCE to make appropriate lanes changes, take approprirate initiative, accelerate when it's needed, de-celerate when that is a better solve for the objective or current situation and just do the damb driving.

I have the feeling that the above answer/solve is being lost in the objective to solve it all (FSD) and it's having a negative impact on just getting the basics right.
The phrase that has been said and repeated by Elon:
"Don't let the perfect prevent the better"
Even though I get frustrated with some features not working perfectly, I do remember when I first drove the model S 2017. I can also remember when I got my 3 on May 2018. What an evolution!!! Can't complain that it is not getting better every day. I can complain that it is not evolving as fast as Elon time.
 

tencate

Top-Contributor
TOO Supporting Member
Joined
Jan 10, 2018
Messages
546
Location
Los Alamos, NM
Tesla Owner
Model 3
Country
Country
Even though I get frustrated with some features not working perfectly, I do remember when I first drove the model S 2017. I can also remember when I got my 3 on May 2018. What an evolution!!! Can't complain that it is not getting better every day. I can complain that it is not evolving as fast as Elon time.
and back in late 2017, things were downright scary! I too am extremely impressed with the progress made. The other day the car needed to get off at an exit and traffic was heavy. Much to my surprise it noticed a gap, signaled, and pulled right in. Me? I would've signaled and waited patiently while some kind soul let me in. Or not.
 

DocScott

Well-known member
TOO Supporting Member
Joined
Mar 6, 2019
Messages
341
Location
Westchester, NY
Tesla Owner
Model 3
Country
Country
I think about this a lot as I drive around. I look at driving scenarios that humans handle with relative ease and question if any automated systems can ever really infer what is going on and ever be able to effectively duplicate the human driver.

A perfect example is a car making a left hand turn about 100 yards in front of you. I can see the gaps in traffic, I can infer when the driver will make the left hand turn and I can be prepared should they decide not to go through a gap in traffic and decide not to make the turn and vacate my lane. All the computer sees is a car stopped 100 yards in front of me. I will maintain speed up to a point where I know/don't know if they are going to take the turn, if they do vacate the lane through an anticipated traffic gap, I can continue at speed and smoothly. The computer just brakes, then accelerates based on an obstacle or not. No inference or contextual knowledge applied and a much rougher trip.
I think this kind of contextual knowledge has gotten better in the last few updates. It's still far, far short of a human, but it's anticipating more things than it used to. For example, consider a car ahead of me, travelling slower than I am, that was changing lanes from the lane to my left to the lane to my right (likely to exit). AP used to be really bothered in that scenario--the slow-moving car crossing in front made it think there was an obstacle, and it would slow down and stay slowed down for quite a while after the car had cleared the lane. But not anymore. Now it correctly recognizes that the car is going to clear out of my lane before I get there.

Similarly, it has no trouble anymore with someone going at my speed who changes in to my lane and stays there, even if the car hasn't left me a lot of room. A context-free system would think "I'm too close!" and slam on the brakes. But AP now does what a good driver would do, and figures the spacing is likely to work itself out soon enough, and that slamming on the brakes is worse for everyone than just allowing the closer follow distance for a few seconds.

I'm not sure FSD will ever be able to match the inference capabilities of a human. But it will have faster reflexes and 360 degree vision. It may not need to match a human's inference ability in order to be a safer driver overall than a human.
 

Mr. Spacely

Well-known member
TOO Supporting Member
Joined
Feb 28, 2019
Messages
365
Location
St. Petersburg, FL
Tesla Owner
Model 3
Country
Country
A perfect example is a car making a left hand turn about 100 yards in front of you. I can see the gaps in traffic, I can infer when the driver will make the left hand turn and I can be prepared should they decide not to go through a gap in traffic and decide not to make the turn and vacate my lane. All the computer sees is a car stopped 100 yards in front of me. I will maintain speed up to a point where I know/don't know if they are going to take the turn, if they do vacate the lane through an anticipated traffic gap, I can continue at speed and smoothly. The computer just brakes, then accelerates based on an obstacle or not. No inference or contextual knowledge applied and a much rougher trip.
What matters the most in FSD is getting me safely from my home to grandmother's house. I really don't care if there is a bit of unnecessary braking. Over time the system will get smoother...
 

MelindaV

☰ > 3
Moderator
Joined
Apr 2, 2016
Messages
10,008
Location
Vancouver, WA
Tesla Owner
Model 3
Country
Country
noticed something slightly different last night driving home when I got to my exit.
With TACC engaged, going from the 55MPH freeway to the long exit offramp, AP/TACC has always incrementally stepped down the speed in 5mph increments on the offramp down to 25MPH. But last night as I came to the bottom of the offramp and to the red light (nobody ahead of me), with the set speed showing as 25MPH, it came to a stop at the light! while in the past, it would slow but if nobody was ahead of me at the light, it would not actually stop.
Now, as a disclaimer, the light is at the end of the offramp with a curve to the left, with a large planter to the right, and there was a car stopped in the left lane - so it is possible it was picking up either the mass of the planter or the car to the left when it decided to stop. but I've come up to this light in the past with the same car to the left but not ahead in my lane and it would have continued thru the intersection.
 
Joined
May 8, 2018
Messages
20
Location
Redmond wa
Tesla Owner
Model 3
Country
Country
What matters the most in FSD is getting me safely from my home to grandmother's house. I really don't care if there is a bit of unnecessary braking. Over time the system will get smoother...
Tell that to my wife who gets car sick due to the unnecessary braking that autopilot currently provides. She'll bear with it for the initial part of a trip and eventually asks that i turn it off and drive manually.
 

Mr. Spacely

Well-known member
TOO Supporting Member
Joined
Feb 28, 2019
Messages
365
Location
St. Petersburg, FL
Tesla Owner
Model 3
Country
Country
Last edited: