All this means is that 10.69.2 will be 
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
FSD Beta 10.69
- Thread starter Buckminster
- Start date
AlanSubie4Life
Efficiency Obsessed Member
Brutal. Happens to everyone. Watch that regen; that’s how you know!
Note the car applied power going into the turn lane. Then maxed out regen (with some brakes I think but would need a bigger screen; the nose of the car dives). There was no way to predict it would need to stop, I guess.
Last edited:
I don't think that's what it means at all. It should be like all the other updates, steady improvement in the system. New neural nets are being implemented, in this case the new occupancy NN for identifying areas in drivable space that should be avoided but not yet registered with the object recognition systems. Such as a pile of dirt near a construction zone.All this means is that 10.69.2 will be
powertoold
Active Member
I kinda know what you might be referring to, but I also can't see anything obvious from the video, other than the lead car stopping aggressively, so fsd beta stopped with more Gs than normal?
Anyway, I won't know until I get 10.69.2.
AlanSubie4Life
Efficiency Obsessed Member
Watch the regen bar. It’s obvious. Don’t even have to look at the speed. That was a rough one.
Regardless of the lead car behavior, note how FSD stopped way behind and didn’t use available space. In any case there was a stop sign with two cars ahead so dawdling was the appropriate behavior here, not accelerating into the turn lane. Coast, then apply a single smooth, high regen stop, stopping at your final target distance from the lead car, and make fine adjustments as that car moves.
No, not more than normal!
Last edited:
Daniel in SD
(supervised)
Yes, it would probably have been smoother had the lead car run the stop sign.
Another thing to notice, in addition to the regen lines, is the predicted path lines that the NN is feeding visualization.
Watch the blue path line on the first left turn. The NN is showing an odd, erratic blue line which gets lighter as the turn approaches, then suddenly the line "snaps" to a gentle left curve. What was that snap? Did one NN take over for another to complete the turn more smoothly?
Watch the blue path line on the first left turn. The NN is showing an odd, erratic blue line which gets lighter as the turn approaches, then suddenly the line "snaps" to a gentle left curve. What was that snap? Did one NN take over for another to complete the turn more smoothly?
Anyone else notice the double fire updates, wider releases and upcoming price increases always occur at Qtr end when revenue is needed.
AlanSubie4Life
Efficiency Obsessed Member
You’d think that, but not the case. The lead car can change the behavior especially if they slow down gradually, but this behavior will occur with no lead car.
Maybe I can start a business being an FSD Shepherd, where I act as the lead car for drivers using FSD. I will call myself The FSD Jesus, and I will guarantee a smooth ride.
The Mercedes cars can line up behind you, as they require a lead car for their L3 system.
powertoold
Active Member
That happens all the time and has happened ever since fsd beta first came out to the OG testers.
I'm going to pay more attention on my visualizations when I'm turning. I always thought mine were smooth on approach and turn.
powertoold
Active Member
FSD beta is designed to drive based on what it sees, so as the road surface that it perceives changes, the path adjusts accordingly.
This can also explain a lot of the false slowdowns. Elon said that 10.69 ("10.13") will begin to drive in areas with no maps, so like a human driving in a foreign neighborhood or trying to find a driveway (and constantly slowing down), FSD beta is doing the same thing, except it's doing it for turns, stop lines, signs, etc. (everything)
AlanSubie4Life
Efficiency Obsessed Member
They can be, but not always. If there is repathing that takes place that the car likes, it often results in a steering wheel jerk. If you watch the videos carefully you will sometimes see this.
Fortunately there is some improvement in this area in 10.69.1.1, and it is less likely to be disturbed by butterflies.
Last edited:
Drove a few hours today on FSD. Most of the time on I-45 south of Dallas on NOA, unfortunately, so not much to report on beta other than I had some cases entering small towns on a 2-lane highway with 70 mph speed limit. FSD beta took over with that speed limit. My car is set for 8% over limit, so beta operated for a short distance at 76 mph. No issues with this, though they were not challenging stretches - basically just lane keeping with no cross traffic.
I did have some cases to monitor the car reaction to speed reductions while on FSD beta. As have been noted, speed reductions still take far too long, though it seemed like they were slightly improved. May just be wishful thinking. I need to set up a controlled test and time one to compare to future versions.
I still see some handoff glitches with entering interstates. Set speed does not update and there is often some lateral jerk as the car gets to the highway. Not just moving over abruptly, but like it enters at the wrong angle then overcorrects back to the left. This was previousy a seamless transition.
One good thing I had not seen before on NOA was while driving in a construction area, with cones, workers and work vehicles on the right, the display popped up a notice like, "Detected construction activity". The car didn't seem to act differently, everything went fine. Just interesting to see the notice.
I've noticed that NOA really loves the left lane on this interstate, even when it's not needing to pass other vehicles. I have it set to move out of the passing lane, but it's more like it's set to move into the passing lane! It would move over when another car came up from behind, but still...
No PB events of note on either NOA or FSD beta, including significant stretches of NOA on 2-lane highway. There were noticeable mirage cars on the visualization at times on NOA, but no reaction from the car to these.
All this driving on 2-lane roads got me wondering when, or if, Tesla will attempt to implement automatic passing of slow traffic on high-speed 2-lane highways? I've never seen the car attempt this and, quite frankly, it's a scary thought!
I did have some cases to monitor the car reaction to speed reductions while on FSD beta. As have been noted, speed reductions still take far too long, though it seemed like they were slightly improved. May just be wishful thinking. I need to set up a controlled test and time one to compare to future versions.
I still see some handoff glitches with entering interstates. Set speed does not update and there is often some lateral jerk as the car gets to the highway. Not just moving over abruptly, but like it enters at the wrong angle then overcorrects back to the left. This was previousy a seamless transition.
One good thing I had not seen before on NOA was while driving in a construction area, with cones, workers and work vehicles on the right, the display popped up a notice like, "Detected construction activity". The car didn't seem to act differently, everything went fine. Just interesting to see the notice.
I've noticed that NOA really loves the left lane on this interstate, even when it's not needing to pass other vehicles. I have it set to move out of the passing lane, but it's more like it's set to move into the passing lane! It would move over when another car came up from behind, but still...
No PB events of note on either NOA or FSD beta, including significant stretches of NOA on 2-lane highway. There were noticeable mirage cars on the visualization at times on NOA, but no reaction from the car to these.
All this driving on 2-lane roads got me wondering when, or if, Tesla will attempt to implement automatic passing of slow traffic on high-speed 2-lane highways? I've never seen the car attempt this and, quite frankly, it's a scary thought!
AlanSubie4Life
Efficiency Obsessed Member
We won’t see this for years, if ever. The visual range and scene comprehension required is extraordinary and it is a difficult task even for a human. It’s L2 so it makes sense to just let the human do this.
I don’t think even L4 vehicles will do it.
willow_hiller
Well-Known Member
Thinking about this critically, this is actually a really good case study in human perception and context clues.
The first (and obvious) context clue in determining which light is ours is one being slightly less visible than the other. This context is probably what most perception networks should attempt to perceive first:
But failing that, and taking this edge case to the extreme, what happens when the lights appear as if they are equal in size? Without distance or shadows, it's literally impossible to know which light is ours in this case:
Oddly enough, I think the next best context clue that we all pick up on (but don't consciously realize we consider) is actually the orientation of the wire on which the lights are. Here's the lights plus the wire as they appeared in the video:
And here's the wire if you were to view it from the perspective of a driver in the perpendicular lane:
So we instinctually know that because the lights are designed to face two adjacent angles in the 4-way intersection, the light that faces the direction that the wire opens must belong to the other lane, and can be ignored.
But, if Tesla has ideas about becoming L4/L5, at some point, they really need to be able to pass the guy doing 45 on a 75 mph road. Certainly not required for L2 or even L3.
Daniel in SD
(supervised)
Nah, the real reason is that it's not necessary for zero intervention drives.
Mine tried passing on the right, I think. Car in front turning left, 2 lane road plus a right turn (also to pass left turns there). I was already for it to slow and stop being the car turning, but it signaled right and went into the extra lane. I thought it would pass on the right, but as it approached the road on the right, decided to kind of turn right there. Got about halfway through the attempted right, but then went straight ahead and stopped, blocking the lane across that road. Kind of hard to explain, but somehow it jumped from passing on the right to trying to turn to the right, and then giving up altogether. (Yes, 69.1.1
Similar threads
- Article
