-
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
I think they just haven’t worked on it yet, videos for 10.69 also show it still doesn’t handle u-turns. Hopefully these are next on the list.
One time I put myself in the outer left turn lane and then engaged FSD. The thing absolutely freaked out when executing the turn - hard breaking and acceleration and crazy steering - it was super pissed that it wasn’t in the inner lane and didn’t like the cars blocking it from hugging the inner turn.
For my neighborhood this is what happens with lanes:
Divided road 4 lanes each direction
Left turn in 2 miles
FSD goes to the right most lane - even into right turn only lanes.
Only about 50 meters from left turn does it start crossing 4 lanes of traffic to get to the left turn - usually not possible with traffic.
In another part of town for an upcoming left it switched from left lane to middle lane 4 times within a quarter mile. No cones or anything on these drives.
Bad map data? Dunno but about 50% of my routes are like this - can’t use FSD at all. Really hoping 10.69 fixes it.
Elon indicated that 10k additional beta users would get this release in about a week. I wonder how they choose those 10k?
I was in the first 'public' group (along with about 1k others) with a perfect safety score who received 10.2 back in October. I would *think* we'd be in this group, but in the past I haven't received beta releases in the earlier waves.
I was in the first 'public' group (along with about 1k others) with a perfect safety score who received 10.2 back in October. I would *think* we'd be in this group, but in the past I haven't received beta releases in the earlier waves.
clydeiii
Member
Not sure where this rumor came from, but I don’t think there was any chance it would be.
willow_hiller
Well-Known Member
clydeiii
Member
Likely a pseudorandom subset of the entire beta testing group. Take a list of all 100k VINs, shuffle the list and push to the first 10k seems most likely. Gives a nice random distribution across diving environments.
TresLA
Member
I would think they’d push to the first 1k then 10k users that give the most feedback (emails and camera icon clips)
And 10.69.2 almost unnoticeably slips from “the week after 10.69.1” to “a few weeks”. Anyone want to make a bet on whether we’ll see it before the end of September?
This is no different than that last few releases. Patience is required.
I'm waiting for the right time to opt in myself, feels a little closer now.
But reading your message, I feel your pain, so I wrote you a little song.
(with apologies to the late great Marty Robbins - I hope you know this tune):
As I navigate on the streets of Laredo
City Streets of Laredo, high-way NoA,
I long to be back in that FSD Beta
That FSD Beta feels so far away.
I don't burn no oil lowly, just drive my Tesla slowly
I drag on that wheel as I'm carried along,
Waitin' for Beta, hope sooner, not later
If my score drops again then I'll know I done wrong.
AlanSubie4Life
Efficiency Obsessed Member
Looks like @Daniel in SD owes me a beer after he gets back online. Haven’t done exact scoring but that is two failures (intervention, and butt sticking out incorrectly, I have not watched the video so this is a minimum but probably just two) in less than 20 attempts, in light traffic.
I’m glad he made the bet apply to any unprotected lefts, otherwise I would be in potential trouble (who wins at 90% is unclear though). I thought that would hurt me but I got lucky.
Will be interesting to see how heavy traffic goes.
Overall it looks like a substantial improvement in these videos. Lots of intriguing incremental improvements.
Whole Mars video, 6:50
Steering wheel: Jerks back and forth in completely unnatural manner.
Whole Mars: That was good.
Lol.
Still looks like we’re gonna be mashing that accelerator all the way to the stop line, though hard to tell exactly how much of an issue there is in the videos.
Two interventions in Whole Mars video (signaling).
Last edited:
AlanSubie4Life
Efficiency Obsessed Member
I disagree. It looks like a big improvement in fundamental capability here. I expected huge improvements in capability, with a significant number of failures still. Looks like that is what we are going to get, though we’ll see how it copes with heavy traffic (that is where it can get tricky of course). I’m hopeful they’ll be able to get at least 80% success rate in heavy traffic based on what we see so far. Might even get close to 90%!
The weird inconsistencies and inconsistent posing and stuff is likely going to persist for a while (resulting in failures), and I am not sure what is the root cause of that, so that could be a long road for all I know.
EDIT: I watched the failure in Chuck’s video (had only listened before). That was pretty egregious, and it does make me worry about perception. It’s hard to know why it went (or at least did not aggressively commit - there was plenty of time) in that situation. As I have said before, if they have the aggression slider that should be able to be slid up a bit and quickly solve that issue. The car clearly showed up on the perception, so seems like not a perception issue, so it just needs to be faster. Increase assertiveness. It may be that with current slider settings that the failure rate is quite high in heavy traffic, if this sort of failure is due to insufficient assertiveness. We’ll see.
Watching in more detail, it is really unclear why it failed. It did not stop on the creep and just rolled in spite of traffic about 5 seconds away (this is an ample gap of course so the go decision is fine). And of course because it is so slow on maneuvering it failed. There was definitely time but it just dawdled. We’ve seen this before of course.
Last edited:
But 90% succes rate is terrible when driving a car. We would all be dead if we had that.
AlanSubie4Life
Efficiency Obsessed Member
Agreed of course. But it’s obviously unreasonable to expect that here; got at least a few years of development still, assuming the hardware has the capability (I’m assuming it would cede control to the driver if it encounters situations where it is not possible with current sensors and other hardware).
Last edited:
45thParallel
Member
I was hoping for at least 2022.20 not that 2022.24 is rolling out. Giving 2022.16 after more than 2 months is disappointing.
willow_hiller
Well-Known Member
Gotta nail that first 9 before you can start the march of 9s.
I’m also waiting to see Chuck do a video in heavier traffic but the video this morning looked very encouraging.
I’m not necessarily concerned yet about the one attempt where the car left a bit of tail hanging out in the median. People do that all the time if there is no traffic coming from that direction in order to allow a more comfortable gap in front from the oncoming traffic. I think we need to see if it scoots up fully into the median in that scenario when traffic begins to approach from lanes behind it and it can’t yet go into the new lanes. I wouldn’t call that attempt a fail myself.
The only real fail I saw was that one attempt where it should have waited before initiating the UPL but then should have been more aggressive about moving into the median once it decided to go for it anyway due to the oncoming car.
I’m grading this as pass vs no pass. It obviously still needs continual refinement on control smoothness even though this looks like a good step forward.
I’m not necessarily concerned yet about the one attempt where the car left a bit of tail hanging out in the median. People do that all the time if there is no traffic coming from that direction in order to allow a more comfortable gap in front from the oncoming traffic. I think we need to see if it scoots up fully into the median in that scenario when traffic begins to approach from lanes behind it and it can’t yet go into the new lanes. I wouldn’t call that attempt a fail myself.
The only real fail I saw was that one attempt where it should have waited before initiating the UPL but then should have been more aggressive about moving into the median once it decided to go for it anyway due to the oncoming car.
I’m grading this as pass vs no pass. It obviously still needs continual refinement on control smoothness even though this looks like a good step forward.
AlanSubie4Life
Efficiency Obsessed Member
We have seen in prior videos (prior builds) it doesn’t and essentially failure on #4 was a demonstration of its blasé attitude to oncoming traffic.
I don’t, except in perhaps some special cases. This is not one of those. You just wait for traffic to clear and then go. Or you proceed all the way to the median. There’s tons of space in this median. I just don’t see the justification in this case.
clydeiii
Member
Tbh it’s unclear he needed to intervene, but he did so out of caution.
Similar threads
- Article
