astrowunder
Member
There should be a little yellow alarm icon in the upper right of your screen- just tap that.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Software Update 2018.39.x (Version 9)
- Thread starter PoitNarf
- Start date
There should be a little yellow alarm icon in the upper right of your screen- just tap that.
My guess is you scheduled the update after they pulled it. So probably just need to wait until they re-launch it. Tesla should really do a better job informing us that the release was pulled after they notified us. Bad, confusing communications leading to a little frustration.
I did it so many times that I am sure I covered all conditions. That I remember right now of the top of my head.
I have done the following:
- Press blinker to start lane change and accelerate remove
- Press blinker to start lane change while accelerating/pedal pressed.
I will try noting better what where the speed limits and the exact order I just know it will do the lane change with pedal pressed and even if let go.
WileyTheMan
Peanut Gallery Member
Yup, this happened to me. I didn't install it before they yanked it (was frantically trying to figure out why I couldn't find the little install icon at the top). Fortunately this thread tipped me off that it wasn't just me seeing this issue.
Got to try V.9 autopilot for a longer trip today (to Tesla service of course lol).
Umm.. pretty scary. Noticeably worse than V.8. Lane changes are way less confisdent. Lane centering was worse (ping ponging). It jumped from the speed i set (was slowing down for stopped vehicles at a stop light) and floored it toward a stopped very large truck at the stop light.
So.. be very careful people. This is not the same autopilot as V.8 ( of course you should always be careful.. but just sharing my experience)
Umm.. pretty scary. Noticeably worse than V.8. Lane changes are way less confisdent. Lane centering was worse (ping ponging). It jumped from the speed i set (was slowing down for stopped vehicles at a stop light) and floored it toward a stopped very large truck at the stop light.
So.. be very careful people. This is not the same autopilot as V.8 ( of course you should always be careful.. but just sharing my experience)
I've seen some reports of v9 being amazingly better, and amazingly worse. I suspect this is why they halted the rollout - something is causing these worse results, and it needs to be ironed out, because it's supposed to be better not worse.
I'm going to guess it has something to do with not-quite-100% identical camera placement (slight misalignment - small enough that us humans would take it for granted) for some cameras combined with the new system using all of them and trying to do so in a supposedly camera agnostic way. My guess is for some combinations the results are bad, borderline dangerous, and they simply didn't have a large enough test fleet to catch it in time.
I'm going to guess it has something to do with not-quite-100% identical camera placement (slight misalignment - small enough that us humans would take it for granted) for some cameras combined with the new system using all of them and trying to do so in a supposedly camera agnostic way. My guess is for some combinations the results are bad, borderline dangerous, and they simply didn't have a large enough test fleet to catch it in time.
the share of Model 3's with 39.7 is up over 70%
As 40.x is starting to trickle out I'm thinking we are just at the next release.
They don't generally continue a wide push once the next release is about ready, they just wait for the next release.
V9 is still going out at a much slower pace to just US. V8 is still going out to everywhere else.
The LTE outage isn't helping much either right now.
Just got my notice that my second update is available (first one failed to install on Saturday so Tesla had to look at the issue and said they would push a new update). Problem is I'm in Anaheim and my car is in Denver. I believe I saw someone post that if you don't install in a certain timeframe after receiving the update, it goes away. Is this true and if so, how long until the update is sent again? I'm gone 'til Friday night so hopefully it will still be there when I get home.
Kermee
It's Not Easy Being Green
The only time it'll go away is if Tesla "pulls" the update due to an issue. Assuming they don't this time, you should be greeted with the update screen as soon as you return to your vehicle.
Ah, neither myself nor my friend with an S had seen an update, and there has been lots of vocal whining about no update across various forums, so I assumed that the rollout was much lower than that.
Of course it could be somehow a disproportionate amount of TeslaFi users got the update and so that stats are not representative, but that seems unlikely.
This wouldn’t surprise me. I see the cars next to me move around a bit on the display but nothing like the various videos people have posted where the other cars are jumping around like crazy and intersecting the Tesla. I feel like AP performance has improved - much less abrupt speed changes and no ping-ponging side to side. Also more assertive lane changes.
The display is likely generated from some separate output from the NN's, not a reconstruction of what actual decision making is based on. So all bets are off on just how accurate that is to what the car's decision making sees. If we had what the car saw visualized, it would be overwhelming and nearly impossible to understand, with too much overlapping data.
Homelink should be inside that launcher/app button where energy/web etc. is. It's way too far to the right now. And I will not use the auto-open/close anymore after almost closing the door on the other car many times.
The point being that if some people have poorly calibrated cameras it could explain the difference in what people are seeing on their screens and the complaints about bad AP behavior.
Oh yeah, that's totally possible. But you can have bad AP behavior and funky on screen behavior without the two having a common cause, too.
In my case, v8 always worked very well (at least, for what it was capable of) but I would often see vehicles gyrating around when slow or stopped, and that was just from the front camera. But the car always drove correctly, it was just the visualization losing it's mind.
So it could be the same oddity in v9, multiplied by many cameras. Some have reported the car driving fine despite the unreal on-screen behavior.
Interesting to see how people have wildly different results.
With my 2015 Model S (running AP1 and V7/V8/V8.1) I typically never used AP because I felt like it took more effort to use than simply using TACC. It would do things like diving for the exits, and there was a some instances of truck lust.
With my Model 3 running V9 I decided to let AP do the vast majority of the driving from Snohomish to Portland, and back. I wanted to use the dash cam feature to document every instance in which it messed up. So I could go back to determine the exact location, and what it might have been thinking.
In that 400+ mile journey I only had one case where there was what I considered to be a significant mistake. That was a false braking event that happened as the result of a overhead road that I could easily see why the car thought it was right in front of it because it was right before the freeway dipped under it. This was on 405 going Northbound right about the time I was thinking "hmm. I haven't experienced any phantom braking yet".
Now it wasn't what I'd consider a lot of braking. But, just maybe 5-10 mph lost as I took over. I didn't have anyone close behind me, but it's still a pretty busy freeway. I'm too self conscious to let things run there course.
Here are my conclusions
False Braking -> One event in 400 miles
Lane Changes -> Overly conservative. It seems to give a second or two just to go "hey, guys I'm going to change lanes", and then it executes a perfect driving miss daisy lane change. The kind of lane change I'd do if I was driving a CEO. It also seemed to require an excessive amount of space behind, and ahead. So it was best used in non-aggressive traffic.
Zig-Zagging -> Didn't notice anything significant. Maybe some wobbles here, and there. Probably outdid average human drivers.
Lane Placement -> Still not a fan of the center in all situations
Truck Lust -> None
Diving for the exit -> None
Lane Change over Solid white lane into HOV lane -> Wouldn't let me
Nags -> Excessive. With the Model S I could hold the steering wheel in such a way to all, but eliminate nags. No such luck on the 3 yet. The warning message is easy to see so I didn't have any major issues with it. Just a little annoying. I'll try a different steering wheel setting next time to see if it improves it (I have sport selected, and it's tight like I like it).
Sensation of using it -> Felt a lot more natural than AP1 did. AP1 tended to constantly do micro corrections. AP2 felt more human like.
The 360 degree visualization -> Comically funny at times, but it was pretty awesome at helping situational awareness.
Final conclusion -> Could use some mad max.
Disclaimer -> I naturally intend to take control when things look like they might get dicey. So I can't say what AP would have done when dealing with weird crap that happens on the roads sometimes. So my analysis is really just 90% of driving on the freeway. I also did most of the more difficult lane changes that required tighter tolerances, and higher aggressiveness than it was delivering.
Now I don't know if I'll be converted from a TACC guy to an AP guy, but it was certainly a large step.
With my 2015 Model S (running AP1 and V7/V8/V8.1) I typically never used AP because I felt like it took more effort to use than simply using TACC. It would do things like diving for the exits, and there was a some instances of truck lust.
With my Model 3 running V9 I decided to let AP do the vast majority of the driving from Snohomish to Portland, and back. I wanted to use the dash cam feature to document every instance in which it messed up. So I could go back to determine the exact location, and what it might have been thinking.
In that 400+ mile journey I only had one case where there was what I considered to be a significant mistake. That was a false braking event that happened as the result of a overhead road that I could easily see why the car thought it was right in front of it because it was right before the freeway dipped under it. This was on 405 going Northbound right about the time I was thinking "hmm. I haven't experienced any phantom braking yet".
Now it wasn't what I'd consider a lot of braking. But, just maybe 5-10 mph lost as I took over. I didn't have anyone close behind me, but it's still a pretty busy freeway. I'm too self conscious to let things run there course.
Here are my conclusions
False Braking -> One event in 400 miles
Lane Changes -> Overly conservative. It seems to give a second or two just to go "hey, guys I'm going to change lanes", and then it executes a perfect driving miss daisy lane change. The kind of lane change I'd do if I was driving a CEO. It also seemed to require an excessive amount of space behind, and ahead. So it was best used in non-aggressive traffic.
Zig-Zagging -> Didn't notice anything significant. Maybe some wobbles here, and there. Probably outdid average human drivers.
Lane Placement -> Still not a fan of the center in all situations
Truck Lust -> None
Diving for the exit -> None
Lane Change over Solid white lane into HOV lane -> Wouldn't let me
Nags -> Excessive. With the Model S I could hold the steering wheel in such a way to all, but eliminate nags. No such luck on the 3 yet. The warning message is easy to see so I didn't have any major issues with it. Just a little annoying. I'll try a different steering wheel setting next time to see if it improves it (I have sport selected, and it's tight like I like it).
Sensation of using it -> Felt a lot more natural than AP1 did. AP1 tended to constantly do micro corrections. AP2 felt more human like.
The 360 degree visualization -> Comically funny at times, but it was pretty awesome at helping situational awareness.
Final conclusion -> Could use some mad max.
Disclaimer -> I naturally intend to take control when things look like they might get dicey. So I can't say what AP would have done when dealing with weird crap that happens on the roads sometimes. So my analysis is really just 90% of driving on the freeway. I also did most of the more difficult lane changes that required tighter tolerances, and higher aggressiveness than it was delivering.
Now I don't know if I'll be converted from a TACC guy to an AP guy, but it was certainly a large step.
ℬête Noire
Active Member
Ok, just tried doing this on "the good road" and it did work for me as well. It seems to buck a bit trying to fit into tighter gaps coming up close to a vehicle ahead in current lane and another behind in the target lane to get it to change over into the gap, and ultimately I don't like the safety gray area that gets me into so won't be doing that again, but even that works.
Similar threads
