FSD Beta 10.69

[powertoold]

What I've noticed on 25.2 so far (could be wrong, very brief):

1) Merging onto highways with stuck traffic is bad again, car tries to move too far to the front when the lane is ending
2) Highway adjacent large trucks seem to be more stable on the visualization (not as much shaking around)
3) Brake lights seem to be much more accurate

 

[kabin]

Oops, 10.69.25.2 near miss. You can see FSDb isn't applying enough brake to stop in time if at all so the driver took over. Being a good FSD influencer he spins it that FSDb would have stopped in time.

 

[AlanSubie4Life]

most all (or, let’s just call it all) of the awful issues of 25.1 are still present in 25.2.

This is because they have been focusing nearly all their energy on ensuring excellent initial polish on v11. If you thought v9 was mind blowing, wait for v11.

Oops, 10.69.25.2 near miss. You can see FSDb isn't applying enough brake to stop in time if at all so the driver took over. Being a good FSD influencer he spins it that FSDb would have stopped in time.

What is funny is that the car slows slightly as the car leaves the lane of travel, then accelerates substantially (2mph quite quickly) as the car starts to move back into the lane of travel!

That acceleration is what elicited the “whoa whoa whoa.” And the disengagement. Totally unnatural and worse than just a late brake. It increased vehicle kinetic energy by 33%.

 

[kabin]

This is because they have been focusing nearly all their energy on ensuring excellent initial polish on v11. If you thought v9 was mind blowing, wait for v11.


What is funny is that the car slows slightly as the car leaves the lane of travel, then accelerates substantially (2mph quite quickly) as the car starts to move back into the lane of travel!

That acceleration is what elicited the “whoa whoa whoa.” And the disengagement. Totally unnatural and worse than just a late brake. It increased vehicle kinetic energy by 33%.

It was wiggin-out a minute or so earlier with a series of ridiculous zig-zag lane changes in a short distance and then total confusion in a left turn. Maybe FSDb was running in the weeds. Gotta hope simple crash avoidance wouldn't fall into an 'edge case.'

 

[Chris_1856]

Update 2022.44.30.10 is the first update I've received since getting FSD Beta.

Wow. Wiped my settings, set location back to factory (Fremont). After driving around a while it fixed the location.

Deleted all my location favorites, and homelink. It took about 6 reboots before FSD was in the options so I could turn it on.

So far the only bug I'm seeing is... interior lights no longer turn on automatically (not sure exactly when that used to work... when you went into park, open the door...).

 

[EVNow]

Oops, 10.69.25.2 near miss. You can see FSDb isn't applying enough brake to stop in time if at all so the driver took over. Being a good FSD influencer he spins it that FSDb would have stopped in time.

You are speculating that he is lying. Atleast say it clearly when its your opinion and not fact.

 

[AlanSubie4Life]

You are speculating that he is lying. Atleast say it clearly when its your opinion and not fact.

The statement you quote does not suggest that the car would not have stopped in time. It’s just a statement of what happens in the video: Did a breakaway segment and showed the car was identified as blue and path was light blue, so FSD beta “probably” would have stopped. This is definitely “spin,” because in reality, no one knows. It appeared to me that regen had started at disengagement, after accelerating towards collision, but hard to say if that would have continued.

 

[dflowerz]

Not a good start to 25.2. Engaged it last night and it turned left normally from a four way stop into the proper lane with a traffic light ahead to cross straight across an arterial. Then out of the blue the car started to veer into the right turn only lane. I had to take over to keep it from doing that. Very scary. Had never done that before with prior versions of FSD. So for that part of the route it is much worse. I have had the car get confused too many times to count trying to turn onto poorly marked roads but this is the first time the car tried to do the absolute wrong thing on a very well marked road.

 

[kabin]

You are speculating that he is lying. Atleast say it clearly when its your opinion and not fact.

Whoa, whoa, whoa. Lets stay on the topic.

Anyone can slow the video down and see inadequate braking was applied. That's an objective fact.

 

[EVNow]

Whoa, whoa, whoa. Lets stay on the topic.

Anyone can slow the video down and see inadequate braking was applied. That's an objective fact.

Anyone can "interpret" how ever they want. That is not an objective "fact".

 

[AlanSubie4Life]

That is not an objective "fact".

The vehicle accelerated into the car in its immediate path, before starting to slow down, before FSD was disengaged. None of this is interpreted, unless we want to get really philosophical.

I think it is fair to call that, factually, inadequate braking. Unless the objective is collisions.

 

[Tronguy]

Finally got 25.2 last night: The Evils of the Once-A-Day check, really. Took it for a spin today. Not very heavy traffic.

1. Still slow turning right, but it gets there.
2. On a wide single lane, when turning left, it still swings wide right before making the turn. (When turning left, vaguely hug the center yellow, don't head for the curb on the right. The car always swings to the right on left turns. Who programmed that?)
3. Handled a left turn at a 4-way stop, with another car at the intersection, cleanly, properly, and at about the same speed a human would have done it. Haven't seen that before.

@AlanSubie4Life, a bit ago, described what sure sounded like an underdamped control system on what should've been a smooth stop: Braking, followed by hard, non-regen braking, followed by gassing it, lather, rinse, repeat for the distance it took to stop. Odd, I thought: I hadn't noticed that. And having a linear/digital control theory background, that's typically the kind of thing that I'd've noticed. So, on this 8-mile round-trip, I was looking for that, every time the car came to a halt in traffic, which happened a few times.

Result: Mostly smooth stops. One where a smooth stop went to a bit harder stop. And one, where there was a red light (and one car waiting at it) preceded by a blinking yellow, where the car came to a near halt at the blinking yellow (it's in front of a fire station), then motored very slowly up to the halted car, a couple hundred feet up.

But no oscillatory action. Hmm.. Dunno, we had a contributor to this thread mention that a replacement of the driving computer fixed some FSDb problems he was having, implying that there was something hardware-like wrong with the replaced computer. Or that a computer replacement truly-duly effectively wiped some code that should've been wiped long ago, in the driving computer or elsewhere.

I'll drive some more over the next few days looking for the oscillatory action. But how common is this, anyway?

 

[EVNow]

The vehicle accelerated into the car in its immediate path, before starting to slow down, before FSD was disengaged. None of this is interpreted, unless we want to get really philosophical.

I think it is fair to call that, factually, inadequate braking. Unless the objective is collisions.

Dirty Tesla might disagree - after all that is your interpretation. I've also disengaged multiple times when I felt the slow down was late or inadequate. DO I think everytime car would have crashed - not at all. I'm never sure.

BTW, a few releases back the car wouldn't slow down until quite late on a right turn. I used to disengage so as not to miss the turn. One day I decided to wait - and guess what it did. It not only slowed down but took a nice turn. Now I don't disengage, and it always brakes late and turns.

It is important to distinguish between facts (perceived late slow down) and speculation (Dirty Tesla is saying collision would have been avoided only because he is an influencer).

ps : I'd have 100% braked in that situation. I check my recordings for such occurrences - and I'm never sure whether the car would have stopped. I'll start carefully checking that dark blue / light blue to make better judgements - but will always brake if the car doesn't slow down fast enough. Not only to avoid possible collision but also to not scare the other drivers.

pps : Again, note we are not discussing whether there is enough evidence whether the car would have stopped or not - it is about whether there is enough evidence to claim that Dirty Tesla lied.

 

[scottf200]

Update 2022.44.30.10 is the first update I've received since getting FSD Beta.
Wow. Wiped my settings, set location back to factory (Fremont). After driving around a while it fixed the location.
Deleted all my location favorites, and homelink. It took about 6 reboots before FSD was in the options so I could turn it on.
So far the only bug I'm seeing is... interior lights no longer turn on automatically (not sure exactly when that used to work... when you went into park, open the door...).

Sorry to hear that. I would be annoyed. When I worked in software, upgrading users from older release was complicated. Occasionally they would have to go to an intermediary release (ie. they waited a very long time to upgrade) because some significant changes were done that affected customization.

How long have you been on FSD?
What was your previous version? (2022.36.20 ... 10.69.3.1?)

TeslaFI example of where people were coming from to get to it.

 

[old pilot]

Dirty Tesla might disagree - after all that is your interpretation. I've also disengaged multiple times when I felt the slow down was late or inadequate. DO I think everytime car would have crashed - not at all. I'm never sure.

BTW, a few releases back the car wouldn't slow down until quite late on a right turn. I used to disengage so as not to miss the turn. One day I decided to wait - and guess what it did. It not only slowed down but took a nice turn. Now I don't disengage, and it always brakes late and turns.

It is important to distinguish between facts (perceived late slow down) and speculation (Dirty Tesla is saying collision would have been avoided only because he is an influencer).

I've seen this behaviour on unprotected LT's when traffic is approaching and the car wants to get to the point where the left turn begins so as to be able to turn in front of the oncoming car. I too, at first, would disengage prematurely but learned to recognize the logic behind what was happening.

 

[legendsk]

Got my 2022.44.30.10 update today. Really pleased to see the improvements to Fart Noise, Light Show, Apple Music, etc. Just a little disappointed to see no improvements in actual use as an automobile. Ah, priorities. . . . . . . . .

 

[AlanSubie4Life]

think it is fair to call that, factually, inadequate braking. Unless the objective is collisions.

DO I think everytime car would have crashed - not at all. I'm never sure.

No one said or implied there would have been a collision! We have no way to know!

“Objective is collisions” was included as an objective. Obviously if the objective was to crash you would want to minimize braking. Obviously that is not the objective. So the braking was inadequate since the car is attempting to avoid collisions. Would it have become adequate? Maybe!

 

[AlanSubie4Life]

I'll drive some more over the next few days looking for the oscillatory action. But how common is this, anyway?

Don’t have 25.2 so I cannot say. Maybe it is fixed! This is quite common in 25.1, but it does depend on various conditions. However, to reproduce the scenario more reliably (though sometimes it will still be ok, so try multiple times!):

1) > 50mph on a high speed road.
2) NO lead car, or stopped lead car.
3) Downhill. The more incline the better. I am fortunate that San Diego has tons of very significant hills.
4) Red light.
5) Take video if possible, otherwise this is simply subjective and there is no measurement.

Wild guess of 50% fail rate in that case (but depends on definition of fail).

If you have a lead car slowing gradually it will usually be ok, of course.

I’m not too worried about my hardware 3.0 at the moment, since the behavior can be observed in countless online videos, to one extent or another. There is one where the camera rips free of the mount on the initial braking.

But no oscillatory action.

To clarify, I have only seen it return to POWER one time. That was remarkable. Again, I suspect 55-60mph with a little downhill would make that possible - you’d have to get that extremely high initial brake slam first. Not sure how common in that specific scenario.

But oscillation is possible even with all slowing, no power application, of course.

 

[EVNow]

No one said or implied there would have been a collision! We have no way to know!

Exactly. That is why I highlighted OP's comment as speculation. OP not only implied collision was a "fact" - he also implied DirtyTesla lied.

Being a good FSD influencer he spins it that FSDb would have stopped in time.

Even when, DT says he thinks the car would have stopped because the other car is in blue and the solid blue line turns to pale blue showing planning to stop.

 

[AlanSubie4Life]

was a "fact" - he also implied DirtyTesla lied.

it is about whether there is enough evidence to claim that Dirty Tesla lied.

No one claimed this. Spin is most definitely not lies!!! It is just providing a favorable interpretation. Which is what was done.

You can see FSDb isn't applying enough brake to stop in time if at all

This is also factually correct, in the observation window. It’s not implying that FSDb was going to crash. You can extrapolate from what we see and see that a collision would have occurred, with no changes in stopping force (not sure what the speed at impact would be, since it is hard to tell precisely when FSDb was disengaged even from the blue wheel). Of course such an extrapolation isn’t meaningful, and no one is saying it is. No one knows! Not even Tesla (since they don’t have the logs, presumably). It’s also not 100% clear that the car was reacting at all (“if at all”) in the observation window. The regen showed up a little, but was that due to disengagement? Was it in response to the obstacle? No idea.