MASTER THREAD: 2019.40.2 - FSD AutoSteer Stop Sign Warning and Adjacent Lane Speeds

[scottf200]

OK, guys, chew on this. I hadn't had any pingponging or wandering at all with 2019.40.2.1 BUT I always drive with a small counterweight on the right side of the wheel. 150g, say 5 Oz of BBs in a lil'bag held with velcro at the right cross-member. That's not enough to defeat the nags. I started doing this because it lets me satisfy the hands on wheel torque request by lightly moving my hand the way I hold it on the right. Without the counterweight I was too easily jerking out of AP, and that was just plain unsafe for me. Today, for the obvious test, I removed it. With no counterweight the ride was entirely different - all sorts of drifting in the lane and other problems. After I put it back, everything is fine again. It's a striking difference. Fellow tweaks can try it and report. It's possible that people who haven't seen drifting are unconsciously applying a little pull to the wheel the whole time. So that the "sometimes it's better/worse" reports come from how we hold the wheel?

Constant subtle but noticeable corrections left and right between the lane. It used to drive dead center straight in the middle of the lane, now it constantly swerves left and right and if you hold the wheel just tight enough you feel the constant corrections in both ways.

Ping-Pong: Those are pretty interesting points. With my signature Model X I received a 'paper weight' in the shape of the car. I've experimented with using this to weigh down one side of the steering wheel BUT I ALWAYS have my hands on the wheel - saves time from when AP tries to kill you. I have not noticed ping-ponging (subtle side to side) with 2019.40.2.1.

Herky-Jerky: I *have* noticed herky-jerky speed up/down/up/down and is magnified by following people not on cruise control. I did a controlled test following an electric car on cruise control and saw much less. when I say 'saw' I mean my collected data using ScanMyTesla.

Re: Ping-pong with and without weight --- could be a good test with ScanMyTesla data.

 

[C141medic]

I’m on 40.2.1.
Definitely getting the swaying/ping ponging on autopilot. Feel like I’m on a cruise ship. Wasn’t there prior to this version of the FW. Lane changes on NOA are perfect. What’s strange though is I no longer get the lane change “nags” to a different lane like I used to. Taking exits with NOA enabled is very aggressive. It takes the exit so abruptly that it goes to the far right of the lane, sometimes crossing the right lane marker, and then immediately centers itself. Not as smooth as prior FW versions.

 

[MarcG]

There’s apparently a goodie hidden in 2019.40.2 (and presumably later) for those who want to spend $2,000 for faster acceleration:

https://electrek.co/2019/12/18/tesla-acceleration-boost-upgrade-leaks-model-3/

 

[dandrewk]

I am confused here. Everyone is saying autopilot is causing ping ponging. Do they mean "auto steer"? I can't see how auto pilot by itself could control steering. I could be wrong though...

 

[Knightshade]

I am confused here. Everyone is saying autopilot is causing ping ponging. Do they mean "auto steer"? I can't see how auto pilot by itself could control steering. I could be wrong though...

Autosteer is part of Autopilot.

 

[dandrewk]

Autosteer is part of Autopilot.

Yes I know. I guess I'm getting a bit confused (again) with TACC. Not the same as Autopilot. It's been a long day...

 

[scottf200]

OK, guys, chew on this. I hadn't had any pingponging or wandering at all with 2019.40.2.1

BUT I always drive with a small counterweight on the right side of the wheel. 150g, say 5 Oz of BBs in a lil'bag held with velcro at the right cross-member. That's not enough to defeat the nags. I started doing this because it lets me satisfy the hands on wheel torque request by lightly moving my hand the way I hold it on the right. Without the counterweight I was too easily jerking out of AP, and that was just plain unsafe for me.

Today, for the obvious test, I removed it. With no counterweight the ride was entirely different - all sorts of drifting in the lane and other problems. After I put it back, everything is fine again. It's a striking difference. Fellow tweaks can try it and report.

It's possible that people who haven't seen drifting are unconsciously applying a little pull to the wheel the whole time. So that the "sometimes it's better/worse" reports come from how we hold the wheel?

Ping-Pong: Those are pretty interesting points. With my signature Model X I received a 'paper weight' in the shape of the car. I've experimented with using this to weigh down one side of the steering wheel BUT I ALWAYS have my hands on the wheel - saves time from when AP tries to kill you. I have not noticed ping-ponging (subtle side to side) with 2019.40.2.1.

Herky-Jerky: I *have* noticed herky-jerky speed up/down/up/down and is magnified by following people not on cruise control. I did a controlled test following an electric car on cruise control and saw much less. when I say 'saw' I mean my collected data using ScanMyTesla.

Re: Ping-pong with and without weight --- could be a good test with ScanMyTesla data.

Re: underlined text above

I have a '17 X 100D (non-raven) AP 2.0 --- I did a controlled test for about 1 minutes time with (blue) and without (pink) a weight on steering wheel at 9 o-clock (270 degree). With weight (blue) is consistently above 0 vs pink line.
I used the same stretch of 6 (3+3) lane highway and going the same speed (60 mph). Far right event line is showing where NoA took me on the same exit both time. I liked up the data points to that.
Aside: You can see in the without (pink) where I had to provide torque after getting prompted by my car. See the two 'event' lines.
KEY POINT: is that on my car I do not see any difference for ping-ponging based on the steering angle recorded on ScanMyTesla.

 

[r1200gs4ok]

just got 40.2.1. ........lost use of NOA and no stop sign / red light stoppage....maybe I’m doing something wrong or bad settings....will figure it out tomorrow

 

[murc585]

Unfortunately the update coincided with a transition to rough weather here in Montreal.
The outside lanes on the highway are often covered by gravel and salt, I can see the autopilot visualisation rendering squigly lines more often.
Driving on the highway in the fast lane is typically where I've noticed the left-right ping-pong or swaying.
I think this is mostly due to the car losing sight of the left lane then targeting the midpoint between the right lane and left shoulder and then seeing the line appear again which causes it to reset its target.
We know the AP is more aggressive in this release, especially when it comes to lane changes, it seems as if Tesla allowed faster speed of wheel rotations, perhaps they are more confident with the perception system.
Dirty Tesla (Youtuber) has uploaded videos of the M3 dodging construction barrels with this release, that makes me think the AP is allowed to perform much more aggressive maneuvers than before.

If this is the case, they need to chill out the lane centering a little bit. I see the NoA blue path line jump left/right much more than before.
If it sees 1 out of 2 lane lines, it should maintain the same distance it had from the visible lane line compared to when it last had 2 visible lane lines.

I've noticed this swaying due to intense winds as well, same with passing trucks, especially in congested cases.
The lane offsets due to trucks seem a lot smarter and consistent in this release, I wonder if this is negatively influencing the smoothness of AP too.
This will be a short-lived problem, this is all software. I appreciate the boost in aggressivity but it seems to be switching targets too often which will probably require increased robustness of the perception system to better predict where hidden lane lines should be and a smarter control algorithm that tries to minimize inducing nausea to its passengers and making us look like drunk drivers.

 

[evannole]

After a week and a half with this release, I am mostly happy with it. I haven't noticed any significant ping-pong behavior as others have.

I do have one criticism: as good, confident and immediate the new lane changes are, the car initiates and completes them a bit too abruptly for my tastes. At times it moves so suddenly and forcefully that I feel like a newly minted teenage driver is at the wheel. I would appreciate a smoother, more gradual action, which would be more comfortable for passengers. I really don't want to feel any G-forces when making a simple lane change.

 

[turns2stone]

Apologies in advance as I'm new here and awaiting delivery, but I have a simple question.

What is the difference in the V10 software vs. the 2019.40.2 (et al) designated updates?

 

[Knightshade]

Apologies in advance as I'm new here and awaiting delivery, but I have a simple question.

What is the difference in the V10 software vs. the 2019.40.2 (et al) designated updates?

Tesla software versions long-form read as YEAR.WEEKOFYEAR.ITERATONOFRELEASE.

So for example 2019.40.2 is the SW created in week 40 of 2019, version 2.

Sometimes they'll go another set of #s for sub-releases like 2019.40.2.1 and 2019.40.2.2 if they make small patches.


2019.32.10 is the first release of the V10 software... essentially just means it's a major revision of the system- and it happens roughly annually (though V8 hung around for 2 years)

 

[Radlaw]

Apologies in advance as I'm new here and awaiting delivery, but I have a simple question.

What is the difference in the V10 software vs. the 2019.40.2 (et al) designated updates?

They are the same. Mine reads: v10.1 (2019.40.2.1 ect.) It can be found under "Software" on your screen.

 

[turns2stone]

Tesla software versions long-form read as YEAR.WEEKOFYEAR.ITERATONOFRELEASE.

So for example 2019.40.2 is the SW created in week 40 of 2019, version 2.

Sometimes they'll go another set of #s for sub-releases like 2019.40.2.1 and 2019.40.2.2 if they make small patches.


2019.32.10 is the first release of the V10 software... essentially just means it's a major revision of the system- and it happens roughly annually (though V8 hung around for 2 years)

They are the same. Mine reads: v10.1 (2019.40.2.1 ect.) It can be found under "Software" on your screen.

Thank you both! Since my account is new, I don't think I can rate your post(s) just yet

 

[Fernand]

I have a '17 X 100D (non-raven) AP 2.0 --- I did a controlled test for about 1 minutes time with (blue) and without (pink) a weight on steering wheel at 9 o-clock (270 degree). With weight (blue) is consistently above 0 vs pink line.

Great you tried this. Not sure how to read your data. The corrective yanks aside, It seems like the pink line has wider swings, but it's not a very long sample. If your weight is heavy enough to never nag, and it's pulling the steering that much, then it's a different case perhaps.

Mine being ~ 155 grams (~ 5.5 Oz) on the 3 o'clock cross (as shown in photo), it doesn't eliminate the nags, it just requires less pull to satisfy them.

If anybody wants to reproduce without a scale, that's 2 full tablespoons of Crosman airgun Copperhead 4.5 mm steel BBs. My other weight I've used in the past is ~ 100 grams, I haven't tried this one on 2019.40.2.1 yet.

On my RWD Model 3 I get a very noticeable reduction in wandering or whatever we call it, and the follow behavior in stop & go traffic is fine, maybe that's unrelated. I wouldn't dream of leaving the little weight off -- it's not a placebo.

The lane changes in this version are IMHO great, marvelously more assertive, with strong "insertion acceleration". Some people might prefer the slower previous way, maybe try Mild (not MadMax) lane change, and Chill acceleration?
.

 

[scottf200]

Great you tried this. Not sure how to read your data. The corrective yanks aside, It seems like the pink line has wider swings, but it's not a very long sample. If your weight is heavy enough to never nag, and it's pulling the steering that much, then it's a different case perhaps.

Statistically, I don't see a wider swing. They appear to be close to the same. Close to 10K sample size is pretty accurate from my background in stats. Could be my pre-Raven X with it's hardware vs Raven X/S. Also the Model 3 may be diff.

 

[willow_hiller]

If anybody wants to reproduce without a scale, that's 2 full tablespoons of Crosman airgun Copperhead 4.5 mm steel BBs.

All arguments about nag-defeats aside, are you worried about how attaching a bag of BBs to the steering wheel could be dangerous in a crash? God forbid they ever made it in front of the airbag.

 

[raygduncan]

All the updates until now for me have been at least stable or improved... but 2019.40.2 is really bad. Autopilot is worse than it has ever been with drifting back and forth in the lane on straight stretches, swerving on curves on the freeway (especially coming out of the curve at speed, when it will sometimes actually verge into the next lane before swerving back), and random phantom braking. Sometimes the braking seems to be in response to a car in the next lane, but sometimes no obvious call at all. On freeway stretches where there is a temporary concrete divider right next to the left lane marker (usually construction areas), it often gets WAY too close to the divider for comfort. Lane changes are more assertive, all right, sometimes startling, almost diving into the next lane. If there was a way to back out this update and wait for the next one, I would.

 

[Fernand]

All arguments about nag-defeats aside, are you worried about how attaching a bag of BBs to the steering wheel could be dangerous in a crash? God forbid they ever made it in front of the airbag.

1) Not really, it's for test. My bag is heavily duct taped, BBs likely to stay together, bag can separate from steering wheel. The plan is to cast equivalent weight into resin. Lots of > 150g objects in cabin. In a crash there are more dangerous things happening. Safer perhaps not to drive, but many accidents happen in the home.

2) It's not a nag-defeat. How many times does one need to say it? A small counterweight like this does not prevent nag/alerts.
.

 

[ecarfan]

Moderator note: Please stay on topic for this thread, which is the new features/issues in 2019.40.2.x

Off topic posts are subject to removal.

Thank you.