Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register
This site may earn commission on affiliate links.
Something interesting happened today, I’ve just completed 100 mile drive in torrid, filthy weather conditions on a busy motorway. Because of the poor weather navigate on autopilot was unavailable so I just drove with autopilot engaged and experienced zero phantom breaking and no hesitancy with overtaking other vehicles. I wonder if this autopilot issue relates to navigate autopilot and not general autopilot driving. I have a few hundred miles of motorway driving upcoming this week and will carry out tests with this theory. Would be interesting to have other opinions too.

There is obviously some correlation between bad weather and navigate on autopilot. Maybe the cameras sensitivity increases with navigate in autopilot engaged 🤔
Well forget my comments above Phantom, breaking and hesitation is still evident when engaging normal auto pilot and not just navigate on auto pilot. Sigh.
 
  • Like
Reactions: Mr. Impatient
One of the things that’s very clear is that AP performance is highly variable. You can be sure that for every new release there will be posts saying “wow, AP is so much better with this version, did a journey of 100 miles and no phantom braking”, and as many saying “wow, AP is much worse with this version, did a journey of 100 miles and had phantom braking all the time”. I know I’ve had out/return journeys on the same day in good weather where it was terrible one way and fine the other. Who the hell knows what’s going on. I think the only real conclusion to draw is that vision-only systems like this will never be reliable enough for mainstream level 3+ automated driving.

Personally, in my view I see AP/EAP/FSD as little more than gimmicks. Useful sometimes, but not to be relied upon.
 
  • Like
Reactions: Alwayshopeful
Personally, in my view I see AP/EAP/FSD as little more than gimmicks.
They’ve demonstrated in the USA that the existing system can cope with driving smoothly and safely on their motorways (with some dodgy lane selection admittedly) so the problem isn’t fundamentally unsolvable.

Just wish they’d bring out a Tesla exec and put them in a UK car for 20 minutes on the M1.

Having said that I’m 99% confident that if I fell asleep holding the wheel with Autopilot activated I’d still be alive when I woke up 20 minutes later.
 
They’ve demonstrated in the USA that the existing system can cope with driving smoothly and safely on their motorways (with some dodgy lane selection admittedly) so the problem isn’t fundamentally unsolvable.
Well I’ve been patiently waiting for 7 years now for them to solve the braking issue and it’s FAR worse now than before.
 
  • Like
Reactions: Mr. Impatient
Echoing the comment of @Pagemakers above, I have collared two mobile engineers on visits and asked them about Autopilot, and they both confessed it's absolutely terrible. Oh, right, I said. Have relegated it to the odd demonstration of brokenness in the middle of Southend at 3am, or down unmarked B-roads where it choose a seemingly random side of the carriageway, much to the horror of passengers.

IT IS NEVER EVER GOING TO WORK PROPERLY.
 
The whole thing took a turn for the worse when the big man said “make it CHEAPER by making it vision only”. It’s that simple.
Andrej Karpathy answers some of Lex Fridman's questions on this,

I think when he says "cost" he's referring at least as much to long term engineering cost, distraction from the central AI problem, quality costs that arise from supporting multiple sensors over time and doing "sensor fusion" (esp. when they conflict), and supply chain management than to unit manufacturing cost and support cost. Fridman seems to be thinking in terms of short term costs including data bandwidth.

Karpathy doesn't mention:
  1. Radar, lidar, & sonar measure distance. Cameras don't. (Stereo cameras as in Subarus could do some distance measurement.)
  2. They're trying to solve a previously unsolved problem at production scale and safety-critical production quality. It'd be smart to put extra money and time into adequately solving the problem before doing cost reductions.
 
  • Like
Reactions: squarepeg77
A comparative review. I've been suffering the motorway brake issue, when passing cars on the left. The cars change dark grey in the FSD visualisation, and then my car brakes unnecessarily. It's a 22 MY with std AP with disabled radar and includes USS. I did not experience this issue ever until i upgraded 2023.32.x.

Just had a new 73 MY delivered. Updated to 2023.32.9. Done 150 busy motorway miles, zero issues. Not a single brake incident, nor did any car in the offside lane change dark grey. AP is also more resilient and confident on A/B roads.
 
  • Like
Reactions: MB11
Same update, but a totally opposite experience...
To be clear, I upgraded the 73 MY from 2023.32.6 to 2023.32.9 and have only driven it with the latter version. Done another 150 M1 motorway miles today and again, no brake issues.

My previous 22 MY on the same update had constant brake issues to the point I was driving with my foot hovering over the accelerator. I logged a support incident with tesla for it. Supplied them videos and timestamps of the issue. Really crappy experience but it's clearly not an issue for all cars. Maybe a poll would help assess the % that have been impacted since 2023.32 was released
 
Latest EAP+FSD auto-steering through A1074 construction. So much for drivable space and cone detection.

1698663447496.png
 
  • Like
Reactions: EatsShoots
The last time I raised a service request about phantom braking (December last year) they requested the dashcam footage when chatting on the app but having gone to the trouble of collecting all of that and getting it on a USB stick for them, they were not interested at all when I went to drop off the car at the service centre and the guy there was puzzled at why I was asked for it in the first place.
There's no way they care. Phantom braking happens to everyone who uses AP. They know it. Spending even a minute on reporting it is like me telling the new york transportation department that the road has pot holes. They know. They don't care and/or cannot/will not fix it.
 
  • Like
Reactions: Pagemakers
There's no way they care. Phantom braking happens to everyone who uses AP. They know it. Spending even a minute on reporting it is like me telling the new york transportation department that the road has pot holes. They know. They don't care and/or cannot/will not fix it.
It's hard for the service team to do much but the company will have to solve this problem for FSD to become possible.

Re-enable radar? Put in stereo front cameras?
 
Reading this forum it looks like AP makes the car undrivable, not my experience at all.

I don't know if it depends on specific cars (model?), traffic conditions, people's perception/tolerance but there are a lot of different experiences with Autopilot.
I have had a 2023 M3 SR+ since the end of September and only had a very few episodes of hesitations/braking.

For example yesterday I drove up and down the M5 Bristol-Birmingham and only had a couple of hesitations and no braking.... maybe someone else had dozens of phantom breaking episodes.... who knows?
 
only had a very few episodes of hesitations/braking.

yesterday I drove up and down the M5 Bristol-Birmingham and only had a couple of hesitations and no braking
These two statements don't add up. Saying you've only had a few episodes since September and then going on to say you had a few episodes yesterday suggests that you've had more than just a few episodes since September. Or are you suggesting it's been flawless up until yesterday?
 
  • Like
Reactions: yessuz