FSD Beta 10.69

[Supcom]

First drive with 10.69.2 today. Most of the same issues (staying in lane on curvy two-lane road, school zones, etc.). Didn't get to test the "go around parked car" bug. But now as I approach a left turn with a short left turn lane from a three-lane highway, the car wants to work right at least one lane. Tried about 1 mile out, then again about 1/2 mile out, and then again about 1000 yards out. I cancelled the lane change every time but maybe tomorrow I will let it change lanes and see what the hell it was going to do to make the left turn. Still doesn't engage the left turn signal until it's already slowed down, entered the left turn lane, and almost come to a stop. I guess people in California are a hell of a lot more patient than drivers in Atlanta.

I've seen 10.69.1.1 move right as it approaches a left turn. I let it go on one yesterday when there was no other traffic. Beta moved back left in time to make the turn, but if it was a busy road, there would be no hope for it.

I tested 10.69.2 today on a high speed two-lane S-curve today and it correctly stayed within the lines on left and right curves driving through in both directions. 55 mph road with 35 mph recommended speed. This was much improved over 10.12.2 for me.

 

[willow_hiller]

And I now have FSD after 9 months of trying. Verified it is 20.15 in the app and on my first drive. I believe the teslafi stat showed one vehicle making the same upgrade. Was not me, I don’t subscribe, but I’m not the only one it would seem…

There's one update on Teslascope from 2022.4: 2022.20.15 Official Tesla Software Update | Teslascope

But out of about 3,000 installs across Teslafi and Teslascope, it's clear that you and that one individual are unique. So maybe Tesla made exceptions for you and others that have been waiting exceptionally long times

 

[powertoold]

Basically all my issues with 69.2 are lane choice / behavior related. It needs to know when to cut into large bike lanes and when to stay in X lane to turn left or right.

 

[aronth5]

Did you try it in the same situations where you've used 10.12.2? Or are you using around pedestrians? I feel like it's noticeably smoother back to back on the same type of low traffic suburban roads i drive. I had a couple of wonky behaviors around pedestrians, but seems like that is to be expected

All my drives were on my regular routes so the only variability was traffic. Very little pedestrian interaction.
I did notice some improvements but they were more than offset by the new disengagements 10.69 introduced.
What's interesting is just like 10.12.2 the road where I'm getting onto without a Stop sign is handled better by FSD then roads with Stop signs. FSD simply drives right to the edge of the intersection with very little creeping, stops then decides when to go. I wish FSD did this for Stop sign intersections when the signs are 20+ feet from the actual intersection. The improved creeping helps but it's nearly as good as how FSD handles a T-intersection without a Stop sign.

I don't recall the exact upgrade but after about a week of driving FSD seemed to settle down so I'm hoping that happens again.

 

[AlanSubie4Life]

I guess people in California are a hell of a lot more patient than drivers in Atlanta.

No, probably not. You have to keep in mind you're speaking to a very eclectic & small group of individuals here, who apparently don't mind how their driving looks or when they get to their destination. It's hard to predict how any particular driver will react, of course, but in general, California drivers are assertive and don't take too kindly to abnormally slow driving. Personally, if I get behind someone very slow (and I need to get somewhere), I patiently back way off, and scream at them in my car; no one would ever know unless watching closely. But everyone is different!

I actually think California drivers are generally pretty good compared to other states I've visited, unless it's raining. They're reasonably accustomed to driving fast, and that cuts both ways.

 

[TheProton]

There's one update on Teslascope from 2022.4: 2022.20.15 Official Tesla Software Update | Teslascope

But out of about 3,000 installs across Teslafi and Teslascope, it's clear that you and that one individual are unique. So maybe Tesla made exceptions for you and others that have been waiting exceptionally long times

I actually had a trial of Teslafi I thought expired, but I went in and checked and it is still live, so that is probably me. Well after a lot of frustration it seems Tesla has helped me out this time! Also I was able to see I was actually on 2022.16.0.3 previously. There were probably not too many people on the .16 branch that got locked out, then did not upgrade beyond it. But still pretty happy to have gotten it today! If someone at Tesla took pity on me thank you!

 

[nvx1977]

Nags on FSDb seem more frequent. Either that or they have upped the torque requirement, and I'm not used to it.

I've noticed since FSDb 10.2 the torque requirement was way less compared to regular AP. Wondering if now they are about the same.

Also, you have less time to respond to the nag. A few flashes and you get the boop boop audible warning.

 

[AlanSubie4Life]

Nags on FSDb seem more frequent. Either that or they have upped the torque requirement, and I'm not used to it.

I've noticed since FSDb 10.2 the torque requirement was way less compared to regular AP. Wondering if now they are about the same.

Also, you have less time to respond to the nag. A few flashes and you get the boop boop audible warning.

Yes, I reported this with 10.69.1.1. The torque requirement is definitely increased. Enough that I frequently disengage (never happened before), though I think I might be able to adapt; just have to be a little closer to the threshold.

I've never seen such a high torque requirement even with the old AP.

I haven't got the audible warning yet. I see the initial warning come up usually, and if it starts to do the blue flashing (used to never see it!) I just start really yanking on the wheel and rapidly disengage, sometimes swerving all over the road, if I can't get that to work (often an issue). Gotta get dialed in on this, and just get used to the stalk disengage instead, if this continues to be a problem, I guess.

One thing I was going to try was to apply torque the opposite direction instead of more torque. I wonder if they're trying to defeat Autopilot Buddy nonsense (which would also be challenged by an increased torque requirement). I'll report back once I get enough chances to try it. They really need to use a NN to detect whether the torque on the wheel is human. Might not even be necessary to use an NN; the torque from AP Buddy is probably quite different than a human hand.

I'm ok with it though. Just have to recalibrate with their silly system. Still, I'd prefer to have a system where it just saw that sometimes I was steering the vehicle, rather than the car (another advantage of a system which allows steering without disengagement).

I haven't driven 10.69.2 yet, but it sounds like the same requirements are there.

 

[nvx1977]

Yes, I reported this with 10.69.1.1. The torque requirement is definitely increased. Enough that I frequently disengage (never happened before), though I think I might be able to adapt; just have to be a little closer to the threshold.

I've never seen such a high torque requirement even with the old AP.

I haven't got the audible warning yet. I see the initial warning come up usually, and if it starts to do the blue flashing (used to never see it!) I just start really yanking on the wheel and rapidly disengage, sometimes swerving all over the road, if I can't get that to work (often an issue). Gotta get dialed in on this, and just get used to the stalk disengage instead, if this continues to be a problem, I guess.

I'm ok with it though. Just have to recalibrate with their silly system. Still, I'd prefer to have a system where it just saw that sometimes I was steering the vehicle, rather than the car (another advantage of a system which allows steering without disengagement).

I haven't driven 10.69.2 yet, but it sounds like the same requirements are there.

I didn't feel like I was on the verge of disengaging when I torqued the wheel. Maybe it's dialed back down a bit on .2 vs .1.1.

 

[sleepydoc]

How are you sure the system is "working" (aka processing very slowly) vs. incorrectly determining that it's not safe to go? If you provide no feedback Tesla will assume that it was not in fact safe to go and that will cause training errors.

You don’t, although ’working’ means processing in an appropriate amount of time, so it it’s taking so long that people are honking or it’s stopping to ‘think’ half way through a turn then it’s not working. But that’s the entire point of my post.

I think you might have made a typo somewhere in there...
The only feedback the user can provide is accelerator input, disengagements, and the report button. Or are you saying that the car can detect when it doing something wrong without any user intervention?

It can detect accelerator input, which you just said is feedback.

 

[2101Guy]

Did some additional drives hoping I would see an improvement after the poor drives of this morning. Unfortunately all my drives this afternoon were no better. Just lots of jerkiness and more disengagements than normal. Just hoping 69.2 improves over the next few days otherwise I may need to reduce the percentage I use FSD. (upgraded from 10.12.2)

 

[momo3605]

All my drives were on my regular routes so the only variability was traffic. Very little pedestrian interaction.
I did notice some improvements but they were more than offset by the new disengagements 10.69 introduced.
What's interesting is just like 10.12.2 the road where I'm getting onto without a Stop sign is handled better by FSD then roads with Stop signs. FSD simply drives right to the edge of the intersection with very little creeping, stops then decides when to go. I wish FSD did this for Stop sign intersections when the signs are 20+ feet from the actual intersection. The improved creeping helps but it's nearly as good as how FSD handles a T-intersection without a Stop sign.

I don't recall the exact upgrade but after about a week of driving FSD seemed to settle down so I'm hoping that happens again.

Yea I agree with your theory though. Idk how to explain it. But 10.12.2 was a mess when I first got it, then I stopped using it for like two months, and then I started using it again a couple weeks ago and it was not that bad anymore lol

 

[j0shm1lls]

Interestingly enough, on the way out to get food right now, the FSD beta stack crashed and gave me the red hands take over message. A few seconds later it disappeared and came back as a frozen OG stack display. 30 seconds or so later, we're back on FSD beta and the rest of the drive was fine.

 

[ChaoticNeutral]

10.69.2 still has many issues that 10.12.2 had in regards to basic functionality. Notably roundabouts are still terrible, and it cannot properly execute a U-Turn at a U-Turn only signal. It also still attempts to change lanes into a parking only lane before turning onto my street.

 

[Oil Freedom]

I noticed the dashcam icon is back at the top -- next to the video reporting button. That frees up one of the customizable button spots at the bottom.

Hmmm.... When parked, the new button at the top only seems to turn the red dot in that button on or off. (The other dashcam button I have at the bottom doesn't change when toggling the top button.) When driving, the button isn't shown.

I must be missing something. What's this thing for???

 

[jeremymc7]

10.69.2 still has many issues that 10.12.2 had in regards to basic functionality. Notably roundabouts are still terrible, and it cannot properly execute a U-Turn at a U-Turn only signal. It also still attempts to change lanes into a parking only lane before turning onto my street.

View attachment 851915

I spent a lot of time with roundabouts last night and it constantly wanted to do between 10-15 when the roundabouts are 35-45.

If it wasn’t 2AM and empty that could have been quite dangerous. Guess it’s better than panic stops, especially as one of the main entrances is at the bottom of a steep hill.

 

[buyleonard]

When parked, the new button at the top only seems to turn the red dot in that button on or off.

That is the Sentry Mode selection...

 

[AlanSubie4Life]

I didn't feel like I was on the verge of disengaging when I torqued the wheel. Maybe it's dialed back down a bit on .2 vs .1.1.

I just took my first drive on 10.69.2, and got a couple of torque warnings which were easily dismissed and no blue flashing screen, so perhaps they did dial it back (but I would agree with you somewhat increased from 10.12, because previously I would likely not have gotten any warnings - will have to recalibrate). Very happy about this, anyway, and hopefully they made it harder to defeat. Looking forward to the "false" strike thread to get hopping again. Is that weird?

Insufficient time to compare to 10.69.1.1.

Four reports to Tesla, one disengagement on my first drive of four minutes. It would be nice if it knew the correct speed on Spring Canyon at Sunset Ridge; it seems convinced it is 25mph limit and this definitely seems to be screwing up the ULT behavior there.

Used the accelerator to go around a right turn corner (Frank Daniels onto Sunset Ridge) faster and it jerked the steering. Reported.
Ironically, on my unprotected left turn, it isn't using the center lane at all to get up to speed, which is an error. It's not visualizing the median at all. Also starts out too far to the left (I think maybe better than 10.12 though?), then turns to the right before turning left. It made a pretty bold (though correct) move to go. Reported.
Followed someone too closely (following distance 7), and following time got down to a little over 1 second, and then of course when traffic turned in front of them and the car in front slowed down slightly, the Tesla slowed down. Not a new problem, needs to be fixed. Reported.
Signaled less than 50 feet before a left turn, started signaling about here. Reported. Not a new problem. I don't understand why this is so hard for Tesla; they know where they are going - for the "signal x distance prior to turn" they seem to be using the position of the traffic signal for the distance, which is just weird!
Disengaged going right into the parking lot, has no idea what to do (expected). Didn't report since there was too much going on (should have, definitely needed a report).

Every single turn got a report (or should have), but we have to start somewhere!

Overall the turns and stuff have been fairly smooth and good, as I have previously reported. It was a very short drive so I can't really assess the stopping jerkiness compared to 10.69.1.1. Definitely I had accelerator interventions on the way back, and there was no change on the late signal right turn, and jerky deceleration, off a high-speed road, which I have posted video of earlier.

Compared to prior version 10.12, I would say it's not a huge step forward (definite improvements though - turning and lane positioning markedly better, and hopefully it "has potential"), but compared to nearly a year ago it is so much better!

 

[powertoold]

I've been saying since the very beginning that reporting does nothing.

Tesla neither responds to the most frequent reports nor addresses them in any way in the release notes.

Maybe someone has evidence to the contrary.

 

[momo3605]

Followed someone too closely (following distance 7), and following time got down to a little over 1 second, and then of course when traffic turned in front of them and the car in front slowed down slightly, the Tesla slowed down. Not a new problem, needs to be fixed. Reported.

I’m pretty convinced follow distance and chill/average/assertive settings don’t do a single thing when on FSDb