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

Wiki MASTER THREAD: Actual FSD Beta downloads and experiences

This site may earn commission on affiliate links.
Still the wrong lane in these 2 cases in san Bruno, South San Francisco.

Turning right from eastbound Sharp Park Road onto southbound Skyline Blvd. Should pull into the number two lane because the number three lane is a right turn only into an apartment complex:
image1.jpeg


From the 101N freeway exit to E. Grand Ave., the car needs to be in the number one lane but really likes the number two lane. Only the number one lane turns left onto grand Avenue.

image2.jpeg
 
  • Informative
Reactions: Jeff N
Or it has to make a quick lane change. Going purely by DOT code, that is probably what is correct.
When everyone is going into the correct lane and there is traffic, going into the wrong lane and then attempting to do a quick lane change is more dangerous and entirely unnecessary. I doubt there is anything in DOT code that says turning into incorrect lanes and swerving around is the correct option.
 
  • Like
Reactions: Jeff N
When everyone is going into the correct lane and there is traffic, going into the wrong lane and then attempting to do a quick lane change is more dangerous and entirely unnecessary. I doubt there is anything in DOT code that says turning into incorrect lanes and swerving around is the correct option.
The tricky part is that there are two right turn lanes. There's a similar right turn that I take all the time, two right turn lanes onto a road with two left turn lanes on to the freeway and a straight through lane that nobody takes. Somehow everyone knows what to do even though the marking on the street shows that the rightmost lane should turn into the rightmost lane. I have no idea how you would program a machine to understand it. At one point the marking on the road matched the reality of how people use the road but for some reason they changed it back.
 
When everyone is going into the correct lane and there is traffic, going into the wrong lane and then attempting to do a quick lane change is more dangerous and entirely unnecessary. I doubt there is anything in DOT code that says turning into incorrect lanes and swerving around is the correct option.
Agreed. This is one of the cases where not following the rules might be the safer thing to do.

But most states have code that says you should turn into the closest lane (single turning lane or corresponding turning lane). There was a discussion on this sometime back and I've linked to NY rules and snapshots etc.
 
reporting back after my weekend driving on the roads previously plagued by PB's thanks to tar stripes etc etc.
FSD 10.11.2 was orders of magnitude better than 10.10.2 in handling the tar marks. On the road that were FSD was previously unusable it was now the smoothest drive I can remember with any release of FSD.
It still tries to stop for the flashing double yellow lights that Texas uses, but the drive out and back from camping this weekend was pretty darned good.
Main issues were annoying stop/go hesitancy when driving towards the sun and the fetish the car has with the left lane.
But overall a noticeable improvement.
 
Just did a Vegas trip - from Orange County - about 250 miles each way. Vast majority of the drive is freeway - so AP/NoA the entire trip. The car did amazingly well. I only had a few (about 4) minor phantom braking issues - like 5mph slowdowns and then back up. One more significant slowdown turned out to be the speed limit changed from 70 to 55 - took me a second to notice why the slowdown. I think it was a truck speed limit sign that was caught far away, but then blocked by a semi truck, so the car took it as a speed limit change. Only happened that once, and there were dozens upon dozens of those 55MPH truck speed limit signs during the trip.

I was very impressed with how smooth it went - and made the trip much more enjoyable.
 
Just did a Vegas trip - from Orange County - about 250 miles each way. Vast majority of the drive is freeway - so AP/NoA the entire trip. The car did amazingly well. I only had a few (about 4) minor phantom braking issues - like 5mph slowdowns and then back up. One more significant slowdown turned out to be the speed limit changed from 70 to 55 - took me a second to notice why the slowdown. I think it was a truck speed limit sign that was caught far away, but then blocked by a semi truck, so the car took it as a speed limit change. Only happened that once, and there were dozens upon dozens of those 55MPH truck speed limit signs during the trip.

I was very impressed with how smooth it went - and made the trip much more enjoyable.
I've noticed that as well, in my case the with temporary school speed limit signs that it used to slow down at all the time. Haven't been through any of those while they're flashing/active to know if it obeys them then though.
 
10.11.2 has been a mixed bag for me. I've noticed many small improvements particularly related to turns. However, when 10.11.2 fails, at least for me, the fails are significant, not predictable, and require aggressive take over. The kind of disengagement where if you're new to FSD you feel you're fighting FSD as it seems to resist your steering wheel correction. Afterwards I say "what the heck just happened" except not that politely!

Phantom "slowness" in general is a bit better except for 1 road where this always happens. One side of the road has a chain length fence most of the way and I'm thinking that may have something to with this but other roads have similar fencing without constant phantom slowdowns.

All in all 10.11.2 is just a very inconsistent build for me.
 
  • Like
Reactions: nvx1977
this is the first release where I felt like I could attempt zero-disengagement drives (allowing for accel pedal use here and there). I've done several routes without disengagement (where previously I've never had a zero disengagement drive). So this version feels like the biggest improvement for a 10.x point release.

But to be clear, car still makes mistakes (seems to prefer turn-only lanes instead of just staying put), and I see a lot more phantom turn signals. Lots of slow creeping and hesitations at T-intersections. Still center-biased on unmarked neighborhood roads. Car still stops at rotaries and waits a while before going, even if no one is around, etc.

Biggest improvements have been comfort-based: whimsical hesitations are significantly reduced. Turns are smoother, although still jittery for some turns. Slowing for sharp curves has improved somewhat (less abrupt) but still could be better (earlier and more gradual).
 
this is the first release where I felt like I could attempt zero-disengagement drives (allowing for accel pedal use here and there). I've done several routes without disengagement (where previously I've never had a zero disengagement drive). So this version feels like the biggest improvement for a 10.x point release.

But to be clear, car still makes mistakes (seems to prefer turn-only lanes instead of just staying put), and I see a lot more phantom turn signals. Lots of slow creeping and hesitations at T-intersections. Still center-biased on unmarked neighborhood roads. Car still stops at rotaries and waits a while before going, even if no one is around, etc.

Biggest improvements have been comfort-based: whimsical hesitations are significantly reduced. Turns are smoother, although still jittery for some turns. Slowing for sharp curves has improved somewhat (less abrupt) but still could be better (earlier and more gradual).
It's interesting how consistent the opinions are on 11.2. I'm not sure I've seen this much consistency in a long time - even the inconsistencies are consistent!

I would point out that "allowing for accel pedal use here and there" is still an intervention and more of an intervention than we tend to realize. When the car is hesitating it's because it doesn't know whether it's safe to proceed or not. When you press the accelerator, you've looked at the road, traffic, pedestrians, etc, and decided it's clear to proceed. Essentially, in a half second you did all the work for the car short of turning the steering wheel.

To be sure, in some of these cases the car would have proceeded eventually and we use the accelerator because we don't want to inconvenience other drivers around us, but part of FSD is responding and acting in an appropriate timeframe.
 
It's interesting how consistent the opinions are on 11.2. I'm not sure I've seen this much consistency in a long time - even the inconsistencies are consistent!

I would point out that "allowing for accel pedal use here and there" is still an intervention and more of an intervention than we tend to realize. When the car is hesitating it's because it doesn't know whether it's safe to proceed or not. When you press the accelerator, you've looked at the road, traffic, pedestrians, etc, and decided it's clear to proceed. Essentially, in a half second you did all the work for the car short of turning the steering wheel.

To be sure, in some of these cases the car would have proceeded eventually and we use the accelerator because we don't want to inconvenience other drivers around us, but part of FSD is responding and acting in an appropriate timeframe.
While it is true that sometimes I press the accelerator so as not to inconvenience those behind me, more often it is for safety reasons. It often is doing that hesitation thing and slowly proceeding in to the lane with oncoming traffic. I press the pedal to get it going before the oncoming is too close.
 
It's interesting how consistent the opinions are on 11.2. I'm not sure I've seen this much consistency in a long time - even the inconsistencies are consistent!

I would point out that "allowing for accel pedal use here and there" is still an intervention and more of an intervention than we tend to realize. When the car is hesitating it's because it doesn't know whether it's safe to proceed or not. When you press the accelerator, you've looked at the road, traffic, pedestrians, etc, and decided it's clear to proceed. Essentially, in a half second you did all the work for the car short of turning the steering wheel.

To be sure, in some of these cases the car would have proceeded eventually and we use the accelerator because we don't want to inconvenience other drivers around us, but part of FSD is responding and acting in an appropriate timeframe.

It's been somewhat of a convention, at least in the old FSD beta videos I used to follow, that interventions and disengagements were tallied separately. So when I say zero-disengagement drives, it's definitely an easier bar for the car than a zero-intervention drive. Plus, I tweak the speed setting with the scroll wheel all the time, and while that one might be less impactful from a "doing the work for the car" perspective, it still means we're a long way away from a comfortable drive, for in-cabin and other cars.

I did note in one of your previous posts (I think it was you?) about patience. What I've found with 10.11.2 is that the car often does do the right thing after uncomfortable hesitations, whereas that was less guaranteed in previous versions. It's for this reason I felt like a zero-disengagement drive was possible. I should also note that for many of these drives, it was past peak traffic hours, like 8pm, so I was more willing to let the car be awkward (less chance of annoying someone).
 
Just did a Vegas trip - from Orange County - about 250 miles each way. Vast majority of the drive is freeway - so AP/NoA the entire trip. The car did amazingly well. I only had a few (about 4) minor phantom braking issues - like 5mph slowdowns and then back up. One more significant slowdown turned out to be the speed limit changed from 70 to 55 - took me a second to notice why the slowdown. I think it was a truck speed limit sign that was caught far away, but then blocked by a semi truck, so the car took it as a speed limit change. Only happened that once, and there were dozens upon dozens of those 55MPH truck speed limit signs during the trip.

I was very impressed with how smooth it went - and made the trip much more enjoyable.
I'm also pretty impressed with NoA except for the tendency to move to a faster lane when a car is going to pass me within a few seconds. I often use my blinker to cancel the lane change. Then there are times when NoA takes too long to make a decision to change lanes which then causes canceling the lane change since cars have then caught up to me.
 
While overall 10.11.2 has been a mild improvement I have several LARGE steps back that makes it impossible to not have a disengagement drive. Here is an example that started with 10.11.2 and is an easy right turn. today it even had a lead truck to follow. Every time it turns on right signal and then veers off to the left. This is a dangerous place to do that with 4 lanes in one direction. I have been honked at twice. I may try and go early in the morning and see if I can catch it with no traffic and let it play out.

 
While overall 10.11.2 has been a mild improvement I have several LARGE steps back that makes it impossible to not have a disengagement drive. Here is an example that started with 10.11.2 and is an easy right turn. today it even had a lead truck to follow. Every time it turns on right signal and then veers off to the left. This is a dangerous place to do that with 4 lanes in one direction. I have been honked at twice. I may try and go early in the morning and see if I can catch it with no traffic and let it play out.

Not enough space to turn ? I see the van in front going over the curb.
 
He didn't hit the curb and there is plenty of room plus Beta has made that same turn close to 100x or more on previous versions without any problem. It has just been since 10.11.2 that it veers to the left and doesn't turn.
Interesting. Shows how difficult it is to get the right path using cost optimization. FSD just thought it was easier to cross over to the left a bit and take the turn. Could also be a local minimum. These problems should get somewhat resolved when the optimization is done in NN.

BTW, looks to me that van went over the curb. The right tyre is higher than the left making the van tilt a bit.

1650390346185.png