I have been driving on .76 for a few days and I just noticed dramatic improvements in stop and go traffic when I changed my car distance to 4. Earlier I was on 2 and the car would abruptly stop a few car lengths behind and aggressively accelerate for a minute and then stop again. Freeway driving has been awesome so far.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
I haven't been in stop and go (yay) with .76 yet. Definitely some steps back (inability to handle wider lanes still but now aggressively hunts, initial snap in lane changes makes passengers uncomfortable, delayed resumption of following lead car at lights, loss of car in front of lead car on IC) but I find a lot more forward on balance (much better lane holding, especially on curves; it seems to see farther and surer (forward radar/vision car image is depicted in the far black shadows of IC and in the correct lanes (mostly); it also appears to be reading signs to me (when signs are obscured it restricts +5, when it sees the sign and sets limit I can go whatever)(my conclusion is that +5 geofence is tied to GPS database and when car sees AP doesn't access the GPS database for geofence on +5); and .76 does not tarry when lane changing.
I'm not sure about the sign reading but I have many anecdotal examples that are building up in my mind. Over time we'll see if its just TomTom being uncannily accurate as compared to Garmin or whether the +5 and sign reading are linked.
I'm not sure about the sign reading but I have many anecdotal examples that are building up in my mind. Over time we'll see if its just TomTom being uncannily accurate as compared to Garmin or whether the +5 and sign reading are linked.
I think it's TomTom's geofencing algorithm causing this. On various onramps and offramps I'm seeing 65mph speed limits (same with driving over bridges sometime) that clearly come from a nearby highway. There's no way the car could've seen the sign at that point.
It just seems like we traded one set of speed limit sign problems for another when switching between these two databases, and that AP1's ability to read speed signs greatly masked some of the fundamental deficiencies in both databases.
Re: speed limits
This has to be one of many critical steps towards FSD. There must be some sort of master plan. I just wish they would let us know if there is some way to improve things in the meantime. I was tired of map quality issues around my house with Waze - so I became an editor and cleaned it up. It feels better to be empowered...
This has to be one of many critical steps towards FSD. There must be some sort of master plan. I just wish they would let us know if there is some way to improve things in the meantime. I was tired of map quality issues around my house with Waze - so I became an editor and cleaned it up. It feels better to be empowered...
Not at all trying to be caustic. SImply angry at Tesla for lying! To me, lying is the worst sin of all (maybe second to murder).
My experience on the new "silky" firmware is much worse than the last. It really bounces around the curves. It dive-and-drives for lanes that widen. It gets really jittery following cars at highway speeds (with distance set to 2), with intermittent but abrupt micro-decelerations. I didn't have that behavior much before. I'm wondering if it has to do with needing time to calibrate again.
I'm experiencing over corrections, ping pong on simple straight lanes where it's slowly left right left right until I stop the over correcting w my hands.
It's impossible in the rain. Worse than 17.17.4 - I'm starting to get frustrated it's 6 months and we can't even bloody drive straight yet.
It's impossible in the rain. Worse than 17.17.4 - I'm starting to get frustrated it's 6 months and we can't even bloody drive straight yet.
FWIW my first 50 or so miles on the highway were pretty rough and then it got a lot smoother on freeways. But that’s strictly for roads with continuous well defined lane markers and no sudden slowdowns. That’s the one environment where this update is leaps better than before and I still think better than AP1.Not at all trying to be caustic. SImply angry at Tesla for lying! To me, lying is the worst sin of all (maybe second to murder).
My experience on the new "silky" firmware is much worse than the last. It really bounces around the curves. It dive-and-drives for lanes that widen. It gets really jittery following cars at highway speeds (with distance set to 2), with intermittent but abrupt micro-decelerations. I didn't have that behavior much before. I'm wondering if it has to do with needing time to calibrate again.
On city roads and with irregular lane markings it’s still a mess.
Well, I decided to drive around for a while. I downloaded the TomTom Go app on iPhone and compared the speed limits. It matched perfectly, except for one 3 mile segment.
Not enough data to be conclusive one way or the other
It is speculated that it is using TomTom for speed limit database. Not in the main screen.
That part I find really strange. I'm seeing, like most reports here, curves consistently handled like a champ. Are the lanes well marked on the roads you're seeing this on?
jlarmstr
Member
From what I have been able to gather .... Google maps is what is pulled via internet to our center console and predominantly the navigation function, but for speed limit info and some nav searching algorithms, it pulls from a separate database -- use to be Garmin, but now it is TomTom, which for me I have noticed several segments of inaccurate speed limit reporting whereas before the update it was pretty much 100% accurate where I drive.
EDIT: And I am not even sure it is google that does the search/route algorithm.... regardless, what it is, it may be replaced soon: Tesla is updating its maps and navigation with open source mapping platforms
SomeJoe7777
Marginally-Known Member
Thanks for reminding us of this, Doug. I think it's easy for everyone to forget that Mobileye has a high level of accountability for touching off this chain of events. Tesla's plan was to run AP2 cars on the EyeQ3 chip side-by-side with TeslaVision until TeslaVision was on par with EyeQ3, then shift over. Then they got the rug pulled out from under them by Mobileye.
Tesla then made the situation worse with the mismanagement of the AP2 customers after that happened. Tesla gave those customers a date that they had no hope of hitting.
You know what Tesla should have done? They should have offered a special deal for about 5,000 customers who currently had AP1 cars. The deal would have been something along the lines of "pay us a small fee now to become a beta tester, get a discount larger than that fee on an AP2 car later". In those cars, they should have then installed a TeslaVision board, and then gathered data from EyeQ3 and TeslaVision running side-by-side. This would have sped up the development cycle of TeslaVision and would only have cost them the cost of the TeslaVision board, labor to install it, and the difference between the discount and the fee.
mach.89
Member
On the same small stretch of Lincoln blvd (PCH) in Marina Del Rey, I've tried 17.26.76 where I've used my AP2 many times before (and even my previous AP1 before that), and I'm not so certain that had I not intervened I wouldn't have rear ended a stopped car at a light maybe 3 or 4 different times since the update. And each time there's a huge gap of no cars in front of me before the stopped car, I have tried setting TAC on farthest following distance (7) the most recent time (usually have it at 4 or 5), and it's as if AP isn't seeing the car stopped in front of me till the last second, i didn't even see it on the instrument cluster till last second. Pretty disconcerting. I'm a fanboy and hate to write something negative but this has me disappointed. I agree with everyone that high speed lane keeping is better, approaching "pretty smooth", not sure about "silky".
jlarmstr
Member
It works on my X (have not tried it on my S), it is just very very subtle. After delivery of my X, I was noticing this weird vibration, like I drove over some of those shoulder warning bumps (but about 1/10th the intensity). It happened randomly and it took a few times until I correlated it to driving across the lines within the lane, then realized that lane departure was on. I can see how you may think it is not working.
Last edited:
Definitely not the 'silky smooth' update I was expecting. In fact, so far, a regression:
- like others, weird stuff going with speed limits - this morning starting speeding to highway speeds in a 40mph zone as i drove under an underpass
- seems *more* uncertain on the highway i travel on most frequently - so much so, i had to turn it off because my wife was afraid I'd look like a drunk driver
- still TERRIBLE in stop and go. Accelerates and decelerates far too quickly i'm afraid the car behind me is going to rear end me. Note - its ok in slow traffic , it's literally in stop and go where it shows its 'inhuman' characteristics
- has not fixed any of the quirks that i'm very much used to on my drive to work, for example this one spot where it inexplicably dives to the right
Early days i suppose, but so far i haven't noticed a single improvement. Hmm...
- like others, weird stuff going with speed limits - this morning starting speeding to highway speeds in a 40mph zone as i drove under an underpass
- seems *more* uncertain on the highway i travel on most frequently - so much so, i had to turn it off because my wife was afraid I'd look like a drunk driver
- still TERRIBLE in stop and go. Accelerates and decelerates far too quickly i'm afraid the car behind me is going to rear end me. Note - its ok in slow traffic , it's literally in stop and go where it shows its 'inhuman' characteristics
- has not fixed any of the quirks that i'm very much used to on my drive to work, for example this one spot where it inexplicably dives to the right
Early days i suppose, but so far i haven't noticed a single improvement. Hmm...
thehumanwire
Member
I just wanted to say that since my last post last Thursday (shown above), it seems like everything's gotten better. I get the feeling some part of autopilot somewhere is learning and it's not making all of the same scary mistakes it was making almost a week ago. I might've been lucky with perfect traffic/driving circumstances, but it seems to be getting better.
I truly think the current version is capable of self learning although not too many believe me
I had a 17.24.28. The first day was rough. After 100 miles later, things got much better. Then I got 17.24.30, back to square 1. Then again, 50 miles later, it got much much better
Received 17.26.76 several days ago, not yet fully tested it
I am wondering for those who had bad experience with this version, did it get better after several trips or several hundred miles later?
When I said "getting better", I meant it was better on the same road with similar weather and traffic condition
Yes, very well marked yet moderate curves where you would feel some lateral G's if you don't slow down. I guess the best way to describe it is that the AP seems too "literal" in that it tries very hard to exactly follow the curve, but then ends up over-correcting. It seems to depict distant cars earlier in some scenarios, but in other situations not detect a car that intrudes into the lane (doesn't slow down).
Also, I had an instance where driving 80MPH a car changed into an adjacent lane and slowed down, and the car thought the car was still in my lane and slammed on the brakes.
The firmware cannot reliably track adjacent cars.
As mentioned before, I'm hoping it's just a calibration issue. I think on the last firmware upgrade the driving got smoother after a few days of use.