AP swinging back-n-forth when shifting lane (2019.8.4)

[caligula666]

For about 4 months now, model S erratically behaves on most lane change to the left by going in between the two lanes and going left right left right etc. dangerously. You can see on the video AP engaged, blinking left, no cars around except one far ahead, AP initiates lane change, gets in between two lanes (hard to see where I am really because of film angle), then start swerving. As we are getting close to the front car, I cancel AP and resume manually.

Called support, sent videos, and after a pretty erratic process they said they checked the logs and all is fine for them. They want me to go the the center so they can experience it themselves but haven’t had time to doing this as it requires me doing this during work hours. Thoughts? Have you had that?

AP2, started in Novemberish 2018, persisted across multiple software releases

 

[mmmk]

Yes I experience this or something like it. I typically watch it like a hawk on the changes and if it begins to suddenly veer back into the original lane I have to instantly grip the wheel and brace to maintain the lane change and disengage. I received 2019.8.4 this past Tuesday. Trying NOA on a regular commute my GF and I noticed right away NOA AP lane changes were smoother but it only completed half of them. The other half it freaked out like your video or disengaged. This is traveling at about ~70 mph, maybe it happens more on curves. I think it's just a limitation of the hardware or software, hopefully just for now. I've had plenty of lane changes fail on well marked highways with little to no traffic. I have a late 2016 AP 2.0 MS 75D. If you hear of any cause please post it.

 

[Alset4S]

So far I've driven about 5,000 miles using AP and or AP with NoA. There have been an occasional lane change cancellation but nothing like you describe.
Once when I was in the left lane of a two lane highway and the highway widened into three lanes the car had a hard time picking a lane. Only happened one time. The lane markings were very faint so that could have been the cause. That occurred with 2018.50.6. The last few days I've been using 2019.8.5 and am impressed so far.
I'm leaving today on a 1,500 mile road trip and will be using NoA most of the time. We'll see how it goes.

 

[caligula666]

Alright, thanks for the feedback, I'll report back here when I have more information

 

[Alset4S]

75D AP2 MCU1 2019.8.5

Sunday 4/6 completed first leg of my 6 day 1500 mile golf road trip. Left Sunday afternoon on 6 hr. NoA drive to Nashville. Best yet by far. Zero phantom breaking, zero aborted lane changes. Lane changes without confirmation was flawless. Sure there where a couple of times when I would have moved out of the passing lane sooner but probably would have felt the same if you were driving. I did cancel a couple of changes out of the passing lane because I could see it was going to almost immediately then pass the car ahead.
After a round of golf Monday morning we're off to Atlanta. My brother in law was extremely impressed with the decisions NoA made and how smoothly it executed. Tomorrow afternoon I'm going to have him drive to Atlanta so I have the perspective of a passenger.

 

[caligula666]

Yes @Alset4S, it seems clear you don’t experience what I experience, good for you. I’ll update the thread for @mmmk who does.

 

[caligula666]

Alright, so it took 7 months but I got it fixed in the end!

First thing I want to say is that those 7 months have been really bad in terms of service, I've been shipped from SC to SC (three in total), things were not followed-up on, emails not answered, etc. it was a bad experience. C'mon, 7 months of active follow-up...

Now, in the end, as it is usually the case, it depends on one person in one SC making the difference and taking it on themselves to solve it. So thank you to that person, the Tesla brand can thank you.

I'm done with the introduction

The problem was with the forward-left parking sensor: it was working fine in "parking mode", and when the SC was doing a diagnostic, it wasn't returning any error. So the person repairing the car had to spend quite a bit of time in discovery mode, making assumption about "what could go wrong despite nothing going wrong", and ended up with this single sensor as the root cause of everything going wrong. They replaced it and they tell me all is good (I didn't get my car yet, I'll comment on how it works as a follow-up). I've been told that the faulty sensor has been shipped to the USA for further inspection as it should have returned proper diagnostic information (imagine what the same issue could mean in terms of lost discovery/repair time across the fleet). I then wondered if other people mentioning issues with AP might have a similar issue (but maybe showing up differently depending on which sensor is faulty) and all is considered fine because the sensor is not returning any error when being diagnosed. Just an assumption obviously.