-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Firmware 9 in August will start rolling out full self-driving features!!!
- Thread starter diplomat33
- Start date
-
- Tags
- 9.0 Tesla Inc.
Great, so I can use the freeway->freeway navigation to get to your house.
boonedocks
MS LR Blk/Blk 19”
That doesn't make any sense unless they really "hack" version numbering - third (aka patch) number is meant for bugfixes, not for staging of completely new functionality. I'm pretty sure they would use new major version (i.e. 2018.34) for that.
That hasn't always been true. The 2017.50.x update was a dot release that added autowiper support. Similarly, 2018.10.4 had the new neural net with the massive AP2 performance but 2018.10 actually was an update too that didn't have any remarkable Autopilot changes.
I really don't think the version numbers mean anything.
With that said, though, many people on Model3OwnersClub have 2018.32.3 and do not report anything terribly new: Firmware build 2018.32.3 baeb637 (8/27/2018)
Can anyone explain what highway on ramp to off ramp really means? Even AP1 is pretty much this on highways. Sure I have to physically make the turn to exit and be nagged frequently but it is pretty solid on hwy. If V9 eliminated the nags and provided a “handoff” time back to the driver of say 10 seconds ala L3 that would be great, especially if it could also take the appropriate exit. I would even be ok with it not changing lanes if it could be a true level 3. I don’t see a huge advantage to it taking the offramp if my hands are on the wheel anyway, probably just responding to a nag.
Agreed. Just with Tesla's history, nothing seems out of the ordinary... I am thinking that the 2018.28.x has the seed code built in it and anything post is just bugfixes until they are ready to unpackage the dormant files that lay within.
Just seems odd that most of the vehicles got sucked up to the last 2 releases over the past week or so...
Nobody has a concrete explanation of this. To date, other than the few vague comments at shareholders' meetings, we haven't found any leaks from employees or beta program members to describe what that feature consists of.
But based off what that exec at the shareholder's meeting said, it basically means the car will move lanes to take exits on its own. So if you are on I-75 and in 20 miles need to take I-94 West per the navigation, it would literally do that whole maneuver on its own with just the stipulation that you supervise and take over if it's being unsafe.
Honestly I don't expect any changes to nag time. Tesla would not have wasted the time to implement all the new nag behaviors in July if they were just planning on ripping it all out in August/September.
As an aside: Have you peeked at the .28 and .32 releases to see if there's anything slightly interesting? I'm starting to believe from personal experience that there's incremental Autopilot performance differences in these releases compared to .26, and I can't tell if it's map tiles, neural net, or something else.
Yes and nothing interesting.
Thanks!
It does seem to feel consistent with the narrative that attention is probably focused on the v9 early access release and hopefully these wide bugfix releases are a sign that v9 is about to start rolling out at least for EAP?
Sigh, now I sound too optimistic.
whitex
Well-Known Member
No you wouldn't. Without lane changes if the car realizes 3 seconds ahead of time that the lane is ending or merging, you don't have 10 seconds to take over - you crash 7 seconds earier. Slamming on the brakes may not be sufficient as there may not be enough time to stop to give you the full 10 seconds, and even if there was, you'd likely cause an accident from behind. See example of abrupt lane change (both lanes are required to move over by one lane due to construction) below - way less than 10 seconds to react. Without lane changes you'd be in the same situation as this poor Tesla driver.
The big feature may seem to be added in A.B.2 as it did not work correctly in version A.B On the other side, the "big changes" in A.B may not be customer facing (i.e. starting to send logs home). And also the bugfix may not bring customer facing change (i.e. too much memory used).
EDIT: of course they may use completely weird/inane process to number version but this is more less standard across the industry.
That wasn’t the case for the two examples I mentioned. Those features were not available in any form before the dot release. It seems like sometimes when they’re trying to shove a new feature out the door, they’re willing to transplant it on top of a “stable” build week branch.
But of course this is extremely rare. The other 99% of the time, a dot release simply has random fixes that halted rollout of the previous major version.
Who knows how they make these decisions... Release management is an art.
And this is why L3 is hard. Everybody looks at how well it seems to handle "normal" highway driving -- when nothing strange is going on -- and think, "Oh, great, I can read a book while it drives and it will just let me know when something weird happens and I need to take over." Nope, there's no time. It needs to be able to handle the strange stuff itself, and that is 10x harder than handling the normal stuff. And it can't even handle all the normal stuff yet.
We won't get L3 even on the highway without a hardware upgrade. On-ramp to off-ramp for EAP will require hands on the wheel the whole time, which is not really what they implied it would be when they sold us our cars. And even when it initiates a lane change they're going to claim that the driver is responsible for making sure the lane change is safe, which is complete BS. If the car takes action on its own I can't be held responsible for that, unless it gives me plenty of time to review its decision and prevent the action if I think it's unsafe. But if it gives the driver this kind of time the opportunity to make the lane change will have passed, unless there are very few other cars on the road. And also, in 99% of cases it will be less headache just to do the lane change yourself rather than babysit an amateur driver trying to do it and asking you for approval.
I guess I assumed it would recognize when the lane would shift. I understand why they say the last 10% takes 90% of the time. Hard
Thats exactly what I was thinking. For me AP is great on the highway, especially morning commute, but I think people’s expectations of what is coming is not realistic. Now you could also make the argument that accidents are going to happen regardless and AP2 L3 fully developed could be much safer than a human driver. The corner case of being in the passing lane behind another car, and there is a disabled car in that lane and the car in front of you swerved at the last minute and you hit the disabled car is referenced often. The reality is humans probably smash into that car also. If we just look at overall safety and concede that the 10% of issues that will take engineers 90% of the time to figure out aren’t ever going to be fully solved, L3 might prove safer than humans. Still going to be a lot easier to blame a machine for an accident. Thank goodness Elon and posse are a lot smarter than me and can hopefully figure this out.
Ok, a little to much Coolaid here. There's no noticeable A / B testing going on, there is no notion of "shadow" mode that Tesla keeps blowing smoke about ( unless you consider snapshots shadow mode, which I do not ), but I guess that could be debated in another thread. People are reading far to much into git hashes, they are just that. If they are working off of feature branches and merging them into master when they are complete then I'd say sure, but Tesla just doesn't do that. They'll put bit's into place to stage them dormantly as they become ready like Valhalla that @verygreen found like a year or so ago and the vector maps flags and such.
However, there is zilch, none, nada v9 (staging) going on with the current firmware ,it's not happening. If v9 drops to beta testers in the next 2 days, I'll donate $200 to a charity of your choice and if doesn't you donate $200 to st. Judes children's hospital? Deal?
Also, I theorize that v9 will need a new NN, just look at the recent fire truck debacle that's still going on, so there will be nothing small about the on ramp / off ramp functionality. In fact, I would not be shocked if Tesla drops the mame crap in a release soon and delays the v9 crap out another few months to say Oct.
Similar threads
