2020.4.1 incoming

[kkillebrew]

Just a follow up. It straightened itself out overnight and was ready to install this morning and went smooth.

Since the last update in December, my HW3 MS keeps giving camera blinded errors while driving in the rain or at night, never had this before. But it seems very sensitive now and will go away typically after 10-15 minutes. I was hoping this latest update would fix this, but it's still happening. Anyone else keep getting those errors?

Same. In the middle of a bad thunderstorm in the dark, in the middle of nowhere. I stoped and looked - nothing obvious - wiped them all off and the error cleared.

 

[Chili]

+1 to the ping-pong issue. HW2.5 here, though it seems like HW3 isn't affected? It appears the lane centering control system is underdamped as the car not only often takes a up to 5 seconds of ping ponging to "find" the center of the lane but also overshoots and crosses lane lines when making significant adjustments (merges, exits, lane width changes, etc.) This has certainly made for some "exciting" moments on autopilot.

I requested the HW3 upgrade in addition to other work at my next service appointment so I'm hoping this issue can at least be taken care of that way.

 

[FMinMI]

Can anyone comment if ping ponging has been fixed, especially if you have HW2 or HW2.5 (I have read that HW3 wasn't really affected). The ping ponging has been so bad since 2019.40.50.7 that I have not been able to use AP at all. Hoping it's been fixed with this update.

I have a MCU1, HW2.0 FSD MS. Got the update last night and drove it today. I have not noticed the ping pong effect on any of the recent firmwares downloaded (most recent was on 2019.40.2.1 before this new one) to my car except when the curb lane is snow-covered and I am in that lane. I did not go very far today but AP seemed fine for the roads I drove. It still brakes hard in stop and go traffic (not running chill and spacing is set to 3) and takes off quickly once the car in front moves. Too early to comment on phantom braking. I did try several voice commands which worked fine (temp, fan speed, heated seats, radio). Headlights off, glovebox open, and a few others came back 'not implemented yet' (or something like that)...

 

[FMinMI]

Since the last update in December, my HW3 MS keeps giving camera blinded errors while driving in the rain or at night, never had this before. But it seems very sensitive now and will go away typically after 10-15 minutes. I was hoping this latest update would fix this, but it's still happening. Anyone else keep getting those errors?

Last week was the very first time in the year I have owned the car where I got message like "right fender camera blinded or blocked" and then later "left fender camera.." and even "multiple cameras impacted" (don't remember the exact wording). It was raining / misting during the drive but not dark. There was some water spray from vehicles tires in front and on the sides. When it came on, it would not let me do auto lane changes when I turned on the 'flicker' but it did not cancel AP - tracking forward was fine. When we stopped I saw nothing wrong with either camera and chalked it up to water causing a blurry image for the computer to see. But it was the first time this happened, and I was on 2019.40.2.1 at the time (but I had that release for a while and had driven in rain before without any issues). May try some RainX on them and now that 2020.4.1 is loaded will be interesting to see if anything changes (although you indicate it still has an issue).

 

[don123]

Ping-ponging happens if I put too much torque on the steering-wheel. If I apply just enough torque to stop the nagging, the ping-ponging stops.

As a test when it’s ping-ponging, stop applying torque altogether. The ping-ponging should stop immediately.

Low tire pressure or non-factory tires with soft walls might magnify a ping-ponging tendency.

 

[NHK X]

I think the ping-ponging you guys are experiencing are the slight nudges AP has been doing lately to move away when passing or being passed by a large vehicle in another lane.

it happens in different ways. Without cars it’s a rhythmic subtle back and forth weaving. With nearby trucks, it swerves pretty severely, more so at night, and worse off in a delayed fashion and in some cases swings back and forth in the lane.

 

[conman]

it happens in different ways. Without cars it’s a rhythmic subtle back and forth weaving. With nearby trucks, it swerves pretty severely, more so at night, and worse off in a delayed fashion and in some cases swings back and forth in the lane.

Interesting. I've experienced the back and forth when passing a truck, but had no ping-ponging in the 16 hours of interstate driving I did a week ago.

 

[NHK X]

Interesting. I've experienced the back and forth when passing a truck, but had no ping-ponging in the 16 hours of interstate driving I did a week ago.

yeah weird that it affects some and not others. Wonder if it happens more with older cars, I have an MCU1 HW2.5 X this occurs on. Can’t test my 3 as my wife didn’t want any autonomy. On a recent trip from Texas to LA it was super annoying and I had to turn off NOA multiple times.

 

[conman]

yeah weird that it affects some and not others. Wonder if it happens more with older cars, I have an MCU1 HW2.5 X this occurs on. Can’t test my 3 as my wife didn’t want any autonomy. On a recent trip from Texas to LA it was super annoying and I had to turn off NOA multiple times.

Mine's MCU1 HW2 S. Maybe, just maybe, it's 2.5 affected this time?

 

[dgatwood]

I was just like you as well, but now get a decent amount of ping ponging. Occurs in a very subtle way without cars around in a very regular back and forth beat. With big trucks and especially at night NOA becomes unusable as the car makes drastic ping pongs in a delayed fashion after cars pass by.

This all started with a couple updates prior (although I’ve been stuck in the past since I have MCU1). Now I’m a believer in the ping ponging and get to eat my shoe, before I thought people were being overly sensitive...

It hasn't been good since the first 9.0 release. That was when it really took a nosedive stability-wise. On the other hand, other aspects of lane keeping (e.g. ramps) got a lot better in that release. Basically, when it is going slow enough for the autopilot hardware to keep up, it does better in later releases, but at highway speeds, it is constantly oversteering and correcting at nearly every single curve.

It took another turn for the worse somewhere in the neighborhood of 2019.32, I think, or maybe a little before, and it feels like 2019.36 was even worse than that. I was only on 2019.40 for part of a day, and it seemed fine, but I only used it on streets with minimal curvature, so I can't say for sure. 2020.4.1 feels about the same as 2019.36 — possibly slightly worse, and possibly not slowing down enough on some tight curves, but those things could be entirely my imagination. It isn't clearly worse, but it definitely isn't better.

+1 to the ping-pong issue. HW2.5 here, though it seems like HW3 isn't affected? It appears the lane centering control system is underdamped as the car not only often takes a up to 5 seconds of ping ponging to "find" the center of the lane but also overshoots and crosses lane lines when making significant adjustments (merges, exits, lane width changes, etc.) This has certainly made for some "exciting" moments on autopilot.

That's consistent with my experience as well, though I've never seen it cross lane lines except on roads with really tight turns, where it ought to be slowing down way more than it does (e.g. Hecker Pass). I have seen it be on top of the lane lines on CA-17 once in a while, though, which is almost as bad. This usually occurs when the thing randomly freaks out and starts oscillating out of control, a problem that can only be fixed by taking manual control, waiting five seconds, and kicking AP back in. I've only seen that a few times.

I wonder if the extent to which it misbehaves is dependent not only upon which car model and AP HW version is involved, but also on which steering rack version that particular car has. Failing to apply the right damping for a given configuration could explain why it seems to get better for some in the same releases where it gets worse for others.

 

[Penfold]

First drive today with this version (came from 2019.40.2.5) and drive was pretty good, still locks onto slow lane cars when switching from AP and NOA but the main problem is the auto wipers no longer work st all. I always try out auto wipers in versions and the last couple of versions have been usable but on this version it hasn’t wiped at all without overriding so have had to turn it off.

 

[DoctorVenkman]

For me, the ping-ponging was 100% caused by software update 2019.40.50.7. It happens whether I am hands off or trying to counter-steer. I have a late 2018 MX 100D with HW 2.5, and this behavior never happened prior to this update.

It's not just passing trucks, either. I recently drove from LA to SF on the 5, which has about 200 miles of completely flat and straight freeway. When I engaged autopilot on the 5, my car would constantly weave back and forth from the right side of the lane to the left side of the lane. It was so bad that my passengers were getting car sick and I had to disengage autopilot the rest of the trip.

This was on a straight road with no traffic. This never happened prior to 2019.40.50.7. No significant physical changes to my vehicle (i.e. tires). The only reasonable conclusion is that the AP build in the past update was a significant downgrade. Luckily I have the 2020.4.1 update queued and I should be able to test AP this weekend to see if it's better.

 

[Silicon Desert]

I always try out auto wipers in versions and the last couple of versions have been usable but on this version it hasn’t wiped at all without overriding so have had to turn it off.

Well, did you reboot the car after the update? Sometimes, rare though, the auto wiper function, along with the auto headlight dimming, get locked up during the update, and require a reboot. It may or may not be your issue, yet worth a try.

 

[Silicon Desert]

My Raven X started downloading this early in the evening. But it looked stuck in “downloading state”. I had to take the car for a short road trip tonight and I figured it would just pickup where it left off or start over when I get back (which may be a good thing). It started to download, stopped and now it says “connect to WiFi for the software update (which I’ve seen come up numerous times with many versions). It’s on Solid WiFi and I even ran SpeedTest.net and I could see it was Verizon FIOS instead of AT&T that the car uses. Model X apparently does not have a shutdown menu like Model 3. I’ve tried reboots and my phone hotspot. Still stuck.

Hmmm, you reminded me of something similar that happened a few days ago during installation of 2020.4.1. There was no problem downloading, but when the MX went to install the update, it took about 45 minutes to get to 30% and then it stayed at that percentage with a "updating auto pilot" message for about 5 hours. Then it went to 60% and stayed at that percentage for another 6 hours with the message of "Updating Electronic Control Units" and then it finally finished installing.

So 11 hours for this installation which was a surprise. Yea, I'm sure there are many people getting it installed in under an hour, but it left me wondering why this was taking so long on my car when all of my previous updates (about 30 of them) generally would take anywhere from 40 to 50 minutes. I hope that was just a fluke.

 

[K2500Z71]

+1 to the ping-pong issue. HW2.5 here, though it seems like HW3 isn't affected? It appears the lane centering control system is underdamped as the car not only often takes a up to 5 seconds of ping ponging to "find" the center of the lane but also overshoots and crosses lane lines when making significant adjustments (merges, exits, lane width changes, etc.) This has certainly made for some "exciting" moments on autopilot.

I requested the HW3 upgrade in addition to other work at my next service appointment so I'm hoping this issue can at least be taken care of that way.

I occasionally see the “overshoot on merge” behavior as well. It seems to occur when merging into the outside of a curved interstate section. My guess is that the algorithm is using the rear facing cameras to read the lane marking to determine when to end the lane change.

 

[Puddles]

My wipers are working properly after the update. I think I am seeing a greater willingness to change lanes right, although I've only seen it because of "faster lane". I wonder what happened to "changing out of passing lane" and why it had to be banished. I liked it, never had any trouble with it.

I've not seen a pronounced ping-pong, just the usual hunting, which seems unchanged also.

 

[PianoAl]

>>might want to blur location on that pic

Why would you need to do that?

Because of people who look as scary as your avatar.

 

[PianoAl]

The search for text in the owners' manual bug is fixed (the search box used to disappear instantly)!

OTOH, I'm now familiar enough with the car that I don't need it.

OTOH, it's good to know that they actually fix bugs.

 

[PianoAl]

I noticed that the release notes on the car mention things that were present already (e.g. backgammon). Is that because some models didn't have that until now?

 

[kevfrey]

Since the last update in December, my HW3 MS keeps giving camera blinded errors while driving in the rain or at night, never had this before. But it seems very sensitive now and will go away typically after 10-15 minutes. I was hoping this latest update would fix this, but it's still happening. Anyone else keep getting those errors?

Yes - I've had this happen since the December update too. However, to Tesla's credit, the December update was also more clear about an error I actually DID have in the side camera hardware, and Tesla replaced the passenger-side camera (kind of a pain, since I had chrome deleted the car, so I had to re-chrome delete that piece after they replaced it - no biggie, just an hour on a Saturday). But, I've had the warning you are mentioning several times during Janurary, sometimes even in clear weather. I wipe the camera when I get out, and sometimes there is very little obstruction. So, perhaps an over-sensitivity to obstructions, or perhaps some odd machine learning anomalies being detected in road conditions that makes it think it can't "see" temporarily, then it clears up after a bit...