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.
I had another drive with v12.3.1 last night.

Usually, this route needs about 3-4 pre-planned disengagement with V11. Again, V12 was able to solve all of them. But, a new problem appeared other than the auto-speed related issues.

As we turn right, we should have stayed on the current lane since we need to go straight, yet it decides to switch to the left lane and got stuck. Happened twice. Shouldn't nav have enough information to handle this kind of situation?

On the way back though, it was impressive, zero engagement.

This was the first experience with my wife. Although it got the lanes wrong twice, she likes this version more. 😁
Exciting
This code is already outdated with 12.3.2.1 is current as of 3/28
Exciting times for fsd
 
  • Like
Reactions: JohnB007 and Pdubs
FSD will always be supervised in one way or another. Just because you're supervised at work does not mean you aren't working. In fact it ensures you're working even more. Ha ha, poor soles. I'm retired. Now I only have one supervisor.
Then what does "F" in FSD stand for? Supervised Full Self Driving is an oxymoron, isn't it?

That's the point I was making. It looks like with vision only you are achieving a glorified autopilot. For Full Self Driving, you probably need more than just vision. And that is what the discussion has always been about. An autopilot is good and relaxing. A good autopilot even more so. But it's not a robotaxi. Not much more to talk about, really, until Elon goes back to making claims that HW 6 or 7 or whatever will one day finally become a robotaxi. Then we'll revisit the arguments and make our bets.
 
Last edited:
Had weird disengagement on 12.3 where the car was completely stopped waiting for traffic, and while stopped the steering wheel starts rapidly turning to the right for no reason. It got almost up to 120 degrees or so of steering angle before I disengaged FSD. If it drove forward like that it would have veered into a planter box.
I recall someone reported this same behavior, but allowed the car to continue. When it was time to move, it straightened the wheel and proceeded normally.

I also recall that it was observed that it was dangerous in their case because they were sitting in a left turn lane waiting for a green. If they had been rear ended while stopped at the light, the car would have been driven to the right and into through-lanes.
 
  • Informative
Reactions: primedive
Does anyone else have a problem waking their car up from a deep sleep using the app it is just not connecting any ideas?
Yes, I've noticed that as well. It used to take perhaps 10 seconds. Now I think it's up to a minute.

Edit: I wonder if this is related to Tesla's intention of reducing the power consumption of the car when parked. Specifically, for Sentry Mode. At home, my car isn't using Sentry Mode, but a change to Sentry Mode handling might affect how often the car checks for an app wakeup.

Shouldn't nav have enough information to handle this kind of situation?
Set up that route again, and look closely at the path to see if the navigation system is telling the car to get into the wrong lane there. It may just be a glitch in route generation, which I think is Google's bailiwick.
 
Last edited:
Yes, I've noticed that as well. It used to take perhaps 10 seconds. Now I think it's up to a minute.


Set up that route again, and look closely at the path to see if the navigation system is telling the car to get into the wrong lane there. It may just be a glitch in route generation, which I think is Google's bailiwick.
I wonder if this sort of thing is why Waymo is approaching the problem differently than Tesla — focus on narrow geographical regions, throw expensive hardware at the problem, but jump straight to L4. Tesla has to figure out how to solve these problems because there will always be bad map data and bad navigation.
 
How old is your car? If it is still a teenager, it might be harder to wake up. /s

Reasons for not connecting using the app I've come across in the past 3.5 years;

1. Worldwide server failure by tesla

2. LTE network down

3. The car is parked or driving somewhere without cell signal (a good one for us was the the only supercharger in the city when we bought the car - it was many levels down in a parking garage with no cell signal.)

4. Not the latest update of the app

5. Apparently if it is very cold, heavy rain, freezing rain, or any other reason I'd want the car to be open the second I arrive at it with my arms full.

(that last one is just me anthropomorphising my car, assuming it hates me as much as I hate it.)
It might be dreaming about being on a hot date with a Rivian
 
For Full Self Driving, you probably need more than just vision.

Why? Humans don't.

THAT said I do think you still have HW limitations present that humans don't have--- human eyes are inside the cabin and don't get mud/snow/road grime covered with no way to clean-- that needs fixing. Humans can also lean forward and get much better side-views around obstructions at intersections than the fixed B-pillar cams can- that needs fixing.

But I don't see why you need more than sight, given humans drive that way- so actual proper sees-everything vision, that is 360 at all times- ought be all you need (sensor wise) to beat a human on safety.

The amount of compute needed is an entirely different topic from the sensor suite of course.
 
Why? Humans don't.

THAT said I do think you still have HW limitations present that humans don't have--- human eyes are inside the cabin and don't get mud/snow/road grime covered with no way to clean-- that needs fixing. Humans can also lean forward and get much better side-views around obstructions at intersections than the fixed B-pillar cams can- that needs fixing.

But I don't see why you need more than sight, given humans drive that way- so actual proper sees-everything vision, that is 360 at all times- ought be all you need (sensor wise) to beat a human on safety.

The amount of compute needed is an entirely different topic from the sensor suite of course.
Mostly because robotaxis are not touted anymore, and instead it's now this Supervised Full Self Driving, which is a word soup. If Elon still believed he can pull off FSD with just vision he'd continue making those claims.

It was discussed a million times how humans have stereo vision, and can move around within a car, and rely on ultrasonic sensors when parking, etc. Most importantly, Tesla is now selling Supervised Self Driving. Case closed.
 
Then what does "F" in FSD stand for? Supervised Full Self Driving is an oxymoron, isn't it?

That's the point I was making. It looks like with vision only you are achieving a glorified autopilot. For Full Self Driving, you probably need more than just vision. And that is what the discussion has always been about. An autopilot is good and relaxing. A good autopilot even more so. But it's not a robotaxi. Not much more to talk about, really, until Elon goes back to making claims that HW 6 or 7 or whatever will one day finally become a robotaxi. Then we'll revisit the arguments and make our bets.
Faux Self Driving
 
Had weird disengagement on 12.3 where the car was completely stopped waiting for traffic, and while stopped the steering wheel starts rapidly turning to the right for no reason. It got almost up to 120 degrees or so of steering angle before I disengaged FSD. If it drove forward like that it would have veered into a planter box.
I had that before 2 times with 12.2.1 and 1 time with 12.3. It did not move the car though. But we should disengage to report bug.
 
How old is your car? If it is still a teenager, it might be harder to wake up. /s

Reasons for not connecting using the app I've come across in the past 3.5 years;

1. Worldwide server failure by tesla

2. LTE network down

3. The car is parked or driving somewhere without cell signal (a good one for us was the the only supercharger in the city when we bought the car - it was many levels down in a parking garage with no cell signal.)

4. Not the latest update of the app

5. Apparently if it is very cold, heavy rain, freezing rain, or any other reason I'd want the car to be open the second I arrive at it with my arms full.

(that last one is just me anthropomorphising my car, assuming it hates me as much as I hate it.)
When my car sleeps like baby I have to tap multiple times to find its location, statuses,...
Same thing with unplugging the charge port.
 
  • Informative
Reactions: SidetrackedSue
Why? Humans don't.

THAT said I do think you still have HW limitations present that humans don't have--- human eyes are inside the cabin and don't get mud/snow/road grime covered with no way to clean-- that needs fixing. Humans can also lean forward and get much better side-views around obstructions at intersections than the fixed B-pillar cams can- that needs fixing.

But I don't see why you need more than sight, given humans drive that way- so actual proper sees-everything vision, that is 360 at all times- ought be all you need (sensor wise) to beat a human on safety.

The amount of compute needed is an entirely different topic from the sensor suite of course.
Human eyes have a lot of limitations with age, or illness, or tiredness, or drugs and alcohol. I have seen cars turning against traffic direction, hitting street dividers. I got rearended mutiple times because people did not see me. I myself turned to wrong streets too.
 
Last edited:
Tesla has to figure out how to solve these problems because there will always be bad map data and bad navigation.
They have a couple million data collection vehicles out there that could clean up that map data in a hurry. I think a lot of folks are amazed that they haven't been using them in that capacity. We certainly reported enough stuff manually during FSD testing.

I wish I had taken a close look at the map back when the car would move three lanes to the left before a right exit. Was that in the route, or was the control software doing that?
 
Virtually the only interventions I need to do are to increase the speed.
Are these interventions for coasting at a higher speed vs starting sooner (from 0 or creeping) vs keeping motion (instead of allowing 12.x to slow too much) vs other? I tap on the accelerator for all 3 of those but most common would be the coasting where at least 12.3 then maintains the higher speed for at least a little bit.

Sometimes I'm also surprised that it decides by itself to go 40+ in a 30mph such as over bridges, but that's how nearby vehicles are driving too, so then I get a bit concerned about the potential speed trap getting off the bridge. Although at least testing so far, 12.3 also seems to realize it needs to slow back down closer to 30mph.