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

New AutoPilot is horrible after update

This site may earn commission on affiliate links.
Chiming in to agree that this latest recall update is bad, and I regret installing it. I’m getting audible nags even with my hands and on the wheel, never had that before, seems to require even more torque than before. I’ve been turning off autopilot to avoid the risk of another strike. I don’t think this update makes my drives safer, probably the opposite. Not teslas choice what they’re done here, hopefully the regulatory tides turn soon and they can revert. Major downgrade.
 
140 mile trip today in mostly clear dry and sunny conditions using a 2023 MYLR, 2023.44.30.5.1, HW3, no FSD, no USS or radar. Auto steer did generally pretty well, only freaked out at one lane merge, a couple of moments of indecision / charging it's mind with not smooth wheel motion at other lane divisions or merges. Steering is still a little more stuttery, jittery, etc than I'd like. The algorithm needs to look about 1 second further ahead to plan, start the corner sooner, and be a little more consistent/smooth with a change in input.

A couple of completely false lane departure warnings. I have it set to only warn, not intervene, and for good reason. The most hilarious one was when the lane I was in was dividing, I tracked right into the new lane (nowhere near the boundary line but it screamed bloody murder anyway), and the depicted lines on screen only showed the old lane to the left. Complete perception fail by the cameras....

Hand on wheel detection was actually excellent. Very very very few nags, sometimes 5+ minutes between nags with just the weight of my hand or a light resistance to its slow see-sawing of the wheel. When it did nag I'd call it legitimate, as I wasn't truly leaving the weight of my hand on the wheel. I'd still prefer capacitive wheel sensor, but the torque algorithm is getting pretty good.

TACC, though, is still garbage. Completely inconsistent about slowing for curves. When it does decide to slow, it starts to slow way too late, too abruptly, too deeply, and then hunts up and down in target speed with hard accel to hard regen cycling back and forth in less than half a second (even with acceleration set to chill). It's nauseating, it disturbs the cars cornering stance, it's just plain dangerous. Then there's the refusing to pass a car that has completely moved to the exit lane (leaving my lane 100% clear ahead) and is slowing to way below the speed limit..... WTF.

Auto steer and nags are headed the right way in my opinion. TACC needs serious TLC and quickly, as it's absolutely ruining our experience with the car.
 
TACC, though, is still garbage. Completely inconsistent about slowing for curves. When it does decide to slow, it starts to slow way too late, too abruptly, too deeply, and then hunts up and down in target speed with hard accel to hard regen cycling back and forth in less than half a second (even with acceleration set to chill). It's nauseating, it disturbs the cars cornering stance, it's just plain dangerous.
How fast were you going? It doesn't work as well when there is a huge speed differential. The pre-slowing is based on pre-mapped checkpoints (as per manual), so it's not always timed correctly or at the right spot, especially if you were blasting down at high speed. What I do is just adjust the speed down with the scroll wheel when I approach a slower curve. I don't think AP will ever handle all slowing for curves by itself, you still need to use it like the assistive function it is.
Then there's the refusing to pass a car that has completely moved to the exit lane (leaving my lane 100% clear ahead) and is slowing to way below the speed limit..... WTF.
I imagine there is still a open split probably (like there isn't a separate lane line yet)? In general, AP is slow to speed back up after slowing for something. That's what the accelerator override is for. This is also something I don't expect to change. It's safer for the system to go slower than to go faster.
 
The new nag messages are infuriating. I have used Autopilot for 4 years now, I am very attentive and always keep my hands on the wheel. Yet recently it is telling me off for briefly glancing down at the screen to adjust a setting - it is demanding more "slight force" on the steering wheel than ever before, which is causing me to make false inputs to the wheel just to please the system. The loud beeps make you panic and think you are about to hit something, which you then start ignoring, meaning any genuine warnings become a boy-who-cried-wolf situation.

The new updates have made Autopilot less useful and more dangerous in my opinion.
 
How fast were you going?
Limit on 90% of yesterday's drive was either 50, 55, or 60mph, and my default is limit +5. We have very windy twisty roads following the river and up mountain canyons, some well marked with the yellow advisory speed signs (lowest on yesterday's route is 45 IIRC), but most not. Yes, for huge cuts in speed from 60 to advisory of 35, I will manually run the TACC max down early for smoothness purposes, and because I do not trust TACC to look far enough ahead to perceive that deep of a cut is required and get there safely and smoothly.

Again my issue is not _that_ TACC slows for curves, but _how_ it does it. Too late, too deep, too inconsistently, and the nauseating and dangerous hunting up and down in the speed that it started doing sometime around 2023.30.6.


The pre-slowing is based on pre-mapped checkpoints (as per manual),
Perhaps that's how it used to work, but there's a lot of things about its current behavior that make me not believe that's still true. First and foremost, TACC will slow for the outside lanes of curve and _not_ slow for the exact same curve going the opposite direction, now on the inside lanes. The slowing for the outside lanes was again, too late, too deep, and super waffling in its decision of what speed to cut down to. The inside lanes it probably should have come down 2 or 3 mph, but did not come down by any from TACC max speed at all.

Second, if your / the manual's assertion that slowing is based on map data is true, why is it hunting up and down in the speed it wants to go? If this curve is map data designated as a "48 mph curve", why does it dive in at 53, stutter (hard on accel then hard on regen and repeat) down to 51, then stutter back up to 55, stutter back down to 49, and repeat?

Third and most damning piece of evidence for that theory is you can run the same curve in the same direction and get completely different results. -8 reduction one time, -2 the next, no slowing at all the next time you run the same curve.

I strongly believe the hard coding changed to an algorithm or machine learning subsystem for TACC to dynamically read the degree of bend in the curve sometime around or just before 2023.30.6 (and someone forgot to update the manual), and that algorithm/ML needs some serious tuning and especially smoothing work still.

I don't think AP will ever handle all slowing for curves by itself
Then how does the Full Self Driving system, that will supposedly one day be running robotaxis do things? Is it just going to run off the road if the map curve checkpoint data is wrong?

Yes, I know I'm on "basic autopilot", "basic TACC", etc and the stacks weren't the same in the past, but it's supposed to be moving towards merged stack for FSD, Auto steer, and TACC, right?

But let's say your assertion is correct, that it's going to get it wrong some percentage of time due to map data error, perception error, whatever. At that point I honestly would rather have an old dumb cruise control that I knew I had to intervene with 100% of the time, versus the current state of affairs where it works 90% of the time, doesn't do any slowing when it probably should 5% of the time, and makes me want to puke 5% of the time.

I imagine there is still a open split probably (like there isn't a separate lane line yet)? In general, AP is slow to speed back up after slowing for something

Nope, lanes were completely split with full solid line for the new lane many yards behind me, lead car had completely moved into the new lane quite a few seconds ago and was deliberately but gently slowing. And so was TACC. It was exactly the behavior I would expect if the lead car was still ahead of me in my lane, trying to hold following distance and everything. Except the lead car wasn't in my lane anymore. At all. And therefore should not be guiding the adaptive part of adaptive cruise control anymore, because it was no longer my lead car, but the system just never figured out to disregard it. There was nobody behind me so I just watched to see what it would do, down to maybe max -15 a across the next 10 seconds or so.
 
Last edited:
Limit on 90% of yesterday's drive was either 50, 55, or 60mph, and my default is limit +5. We have very windy twisty roads following the river and up mountain canyons, some well marked with the yellow advisory speed signs (lowest on yesterday's route is 45 IIRC), but most not. Yes, for huge cuts in speed from 60 to advisory of 35, I will manually run the TACC max down early for smoothness purposes, and because I do not trust TACC to look far enough ahead to perceive that deep of a cut is required and get there safely and smoothly.

Again my issue is not _that_ TACC slows for curves, but _how_ it does it. Too late, too deep, too inconsistently, and the nauseating and dangerous hunting up and down in the speed that it started doing sometime around 2023.30.6.



Perhaps that's how it used to work, but there's a lot of things about its current behavior that make me not believe that's still true. First and foremost, TACC will slow for the outside lanes of curve and _not_ slow for the exact same curve going the opposite direction, now on the inside lanes. The slowing for the outside lanes was again, too late, too deep, and super waffling in its decision of what speed to cut down to. The inside lanes it probably should have come down 2 or 3 mph, but did not come down by any from TACC max speed at all.

Second, if your / the manual's assertion that slowing is based on map data is true, why is it hunting up and down in the speed it wants to go? If this curve is map data designated as a "48 mph curve", why does it dive in at 53, stutter (hard on accel then hard on regen and repeat) down to 51, then stutter back up to 55, stutter back down to 49, and repeat?

Third and most damning piece of evidence for that theory is you can run the same curve in the same direction and get completely different results. -8 reduction one time, -2 the next, no slowing at all the next time you run the same curve.

I strongly believe the hard coding changed to an algorithm or machine learning subsystem for TACC to dynamically read the degree of bend in the curve sometime around or just before 2023.30.6 (and someone forgot to update the manual), and that algorithm/ML needs some serious tuning and especially smoothing work still.


Then how does the Full Self Driving system, that will supposedly one day be running robotaxis do things? Is it just going to run off the road if the map curve checkpoint data is wrong?

Yes, I know I'm on "basic autopilot", "basic TACC", etc and the stacks weren't the same in the past, but it's supposed to be moving towards merged stack for FSD, Auto steer, and TACC, right?

But let's say your assertion is correct, that it's going to get it wrong some percentage of time due to map data error, perception error, whatever. At that point I honestly would rather have an old dumb cruise control that I knew I had to intervene with 100% of the time, versus the current state of affairs where it works 90% of the time, doesn't do any slowing when it probably should 5% of the time, and makes me want to puke 5% of the time.



Nope, lanes were completely split with full solid line for the new lane many yards behind me, lead car had completely moved into the new lane quite a few seconds ago and was deliberately but gently slowing. And so was TACC. It was exactly the behavior I would expect if the lead car was still ahead of me in my lane, trying to hold following distance and everything. Except the lead car wasn't in my lane anymore. At all. And therefore should not be guiding the adaptive part of adaptive cruise control anymore, because it was no longer my lead car, but the system just never figured out to disregard it. There was nobody behind me so I just watched to see what it would do, down to maybe max -15 a across the next 10 seconds or so.
You are describing the problems old code AP has, which I think Tesla have not been working on for years. I also don't think Tesla have merged the current FSD codes to AP yet. The current FSDbeta code handles winding mountain roads much better.
 
The last update (Christmas Update) definitely nerfed AP and EAP.

After getting the update I noticed AP would not engage with the first try and would always have to try and engage it 2 - 3 times before it would activate.
And most often kept getting the 'Auto Steer unable to activate' message when it would fail.

Then last week went on a road trip to Montreal and there was quite a bit of rain both ways, but it was no more then anything I have driven on with EAP/AP on. And EAP/AP was absolute garbage during this trip, where the cruising speed would constantly drop and sometimes well before the average speed of traffic, stating bad driving conditions.
If the traffic was moving at about 110 km/h (E)AP speed would sometimes drop to 60km/h and it was getting dangerous to drive with AP on. Even though the cameras can see and make out the road lines and vehicles. It was getting to a point where I was getting stressed and anxious with AP on, so I had to disengage it most of the time. I have driven in

I opened a SC request stating all the problems I had and got this response,

" We have reviewed your upcoming appointment and determined that the concern you are experiencing is caused by a known firmware issue, The best thing you can do is continue to connecting your vehicle to Wi-Fi so that it can download and install new firmware updates as they come available. Our engineering department is currently wokring on a fix. Once the fix is available it will be sent out via firmware update. "

Funny enough I got a new code pushed to my car, 2023.44.30.7 today. Will need to see if this is the fix they were talking about.
 
So now there's 7 revisions of the same update??? How many more revisions will Tesla need to get this right?
So glad I'm still holding onto 2023.44.1 :cool:
Nope, there are only 6 different versions of the update. But they aren't all the same, since there appears to be different versions for different hardware. There were only 2 versions that went wide, one for each hardware platform. There is now a second version starting to go wide for one of the platforms.

We don't know if all the different versions were because of bugs, at least some were, but they could have been to add recall support for different AP hardware configurations as well.
 
Limit on 90% of yesterday's drive was either 50, 55, or 60mph, and my default is limit +5. We have very windy twisty roads following the river and up mountain canyons, some well marked with the yellow advisory speed signs (lowest on yesterday's route is 45 IIRC), but most not. Yes, for huge cuts in speed from 60 to advisory of 35, I will manually run the TACC max down early for smoothness purposes, and because I do not trust TACC to look far enough ahead to perceive that deep of a cut is required and get there safely and smoothly.

Again my issue is not _that_ TACC slows for curves, but _how_ it does it. Too late, too deep, too inconsistently, and the nauseating and dangerous hunting up and down in the speed that it started doing sometime around 2023.30.6.



Perhaps that's how it used to work, but there's a lot of things about its current behavior that make me not believe that's still true. First and foremost, TACC will slow for the outside lanes of curve and _not_ slow for the exact same curve going the opposite direction, now on the inside lanes. The slowing for the outside lanes was again, too late, too deep, and super waffling in its decision of what speed to cut down to. The inside lanes it probably should have come down 2 or 3 mph, but did not come down by any from TACC max speed at all.
When you come in the opposite direction, the checkpoint is obviously different. If they were at the same spot it would be wrong since the checkpoint needs to be before the curve. Note I'm talking about checkpoints where the car automatically sets the set speed lower (then restores your previous set speed after the curve is over).
Second, if your / the manual's assertion that slowing is based on map data is true, why is it hunting up and down in the speed it wants to go? If this curve is map data designated as a "48 mph curve", why does it dive in at 53, stutter (hard on accel then hard on regen and repeat) down to 51, then stutter back up to 55, stutter back down to 49, and repeat?

Third and most damning piece of evidence for that theory is you can run the same curve in the same direction and get completely different results. -8 reduction one time, -2 the next, no slowing at all the next time you run the same curve.
Again to clarify, I'm talking about premapped points here the car sets a lower set speed by itself (meaning the set speed on the right actually goes down). For example I have a curve I commonly drive at and it will lower the set speed by itself to 70mph well ahead of the curve at the exact same spot every single time. Then after the curve is over, it will restore the previous set speed (say for example 74mph). It's not a fixed negative speed setting.

As for the car going up and down in speed (while presumably the set speed is not changed), it's probably because you have a set speed set that is too high for the curve. So the car will slow down when it fails to keep itself on the curve, but will do its best to try to speed back up to your set speed.
I strongly believe the hard coding changed to an algorithm or machine learning subsystem for TACC to dynamically read the degree of bend in the curve sometime around or just before 2023.30.6 (and someone forgot to update the manual), and that algorithm/ML needs some serious tuning and especially smoothing work still.
I don't have that version yet, so I can't speak to how that version operates. My experience is on the radar version and also the holiday update version that came right after the switch to Vision 2022.44.25.3.
Then how does the Full Self Driving system, that will supposedly one day be running robotaxis do things? Is it just going to run off the road if the map curve checkpoint data is wrong?

Yes, I know I'm on "basic autopilot", "basic TACC", etc and the stacks weren't the same in the past, but it's supposed to be moving towards merged stack for FSD, Auto steer, and TACC, right?

But let's say your assertion is correct, that it's going to get it wrong some percentage of time due to map data error, perception error, whatever. At that point I honestly would rather have an old dumb cruise control that I knew I had to intervene with 100% of the time, versus the current state of affairs where it works 90% of the time, doesn't do any slowing when it probably should 5% of the time, and makes me want to puke 5% of the time.
Even though there is a lot of talk about "single stack", basically AP largely is its own code. FSD Beta works significantly different and Tesla is not obligated to use that code in normal AP (might not be a good idea either as sometimes FSD has regressions).
Nope, lanes were completely split with full solid line for the new lane many yards behind me, lead car had completely moved into the new lane quite a few seconds ago and was deliberately but gently slowing. And so was TACC. It was exactly the behavior I would expect if the lead car was still ahead of me in my lane, trying to hold following distance and everything. Except the lead car wasn't in my lane anymore. At all. And therefore should not be guiding the adaptive part of adaptive cruise control anymore, because it was no longer my lead car, but the system just never figured out to disregard it. There was nobody behind me so I just watched to see what it would do, down to maybe max -15 a across the next 10 seconds or so.
Is it increasing speed or just holding the speed? Was the car still showing up as the target car (darker)? It's possible it's anticipating the car being incisive about the chance and possibly going back into your lane. If instead it was increasing speed, just slowly, that seems to match how it typically operates after slowing down for a car in front (I use accelerator override to overcome that).
 
Got the .8 update and installed it.
It estimated the install time at 1 hr 54 minutes? Each previous update has generally taken 20 - 30 minutes. This one took around 30 minutes.

Dry wipes are back again.
Still insists on driving in left (oncoming traffic), lane on 2 lane road.
If disengage and force it to right lane, then when finally meets oncoming car, slams on regen and brakes and slides to full stop, crosswise in right hand lane with screaming alerts and flashing red wheel of death.

Alternate routes gone, nav can only go on original and FSD ignores nav guidance and goes to opposite turn lane then collapses in the turn and turns control over with screaming alerts and flashing red.

Still have to turn everything possible off and just drive it as if it were something from 1950.
 
Got the .8 update and installed it.
It estimated the install time at 1 hr 54 minutes? Each previous update has generally taken 20 - 30 minutes. This one took around 30 minutes.

Dry wipes are back again.
Still insists on driving in left (oncoming traffic), lane on 2 lane road.
If disengage and force it to right lane, then when finally meets oncoming car, slams on regen and brakes and slides to full stop, crosswise in right hand lane with screaming alerts and flashing red wheel of death.

Alternate routes gone, nav can only go on original and FSD ignores nav guidance and goes to opposite turn lane then collapses in the turn and turns control over with screaming alerts and flashing red.

Still have to turn everything possible off and just drive it as if it were something from 1950.
Something is very wrong with your car.
 
  • Funny
Reactions: David99
This new recall update has made FSD AutoPilot basically unusable.

It has a 50/50 chance of triggering a hands-on-wheel audible warning within one second after engaging AutoPilot, which results in me immediately disengaging AutoPilot to avoid the risk of getting a strike.

Audible nags trigger before I see the visible nag 100% of the time now.

Nags occur constantly with both hands on wheel at a rate that is bad enough to be a severe distraction from driving.

I think this may be the update that makes AutoPilot so bad that I stop using it entirely.
 
...

After getting way more bags than I’ve ever seen since 2018, AP finally shutdown and said no more AP available for rest of drive.

I find this absolutely ridiculous.

...

Were my hands off the steering wheel for more than 5-15 seconds ? Yes

Was I looking down at my phone occasionally sending a text? Yes a few times.

And I would bet that 90% of Tesla AP drivers do the above , so it’s not out of the ordinary.

Several times, when I saw the flashing blue nag, I reached for steering wheel, but twice I was 1-2 seconds too late, causing RED flashing - which is VERY rare for me prior to update. I never let it go in the RED flashing state 99% of my drives.

2 times, I was going around a curve at 80 mph and I got the nag - I applied gentle torque bc I was in the apex of a turn, not wanting to crash, I was slow and gentle with my hands on the wheel —- result? I got AP flashing red and it disengaged for rest of drive

Wtf ? AP is the reason I purchased a Tesla.
Your (and most other's complacent) behavior while operating AP/FSD is problematic and is exactly why regressive changes were made to the AP and FSD feature set. I hate it as well. But as the saying goes... it only takes a couple of bad apples to ruin the batch.

Checkout this video which goes over several close calls with FSD and actually talks about the NHTSA findings that caused this regression of the feature. He makes some good points: