Disengagements mean almost nothing without scenario data. 1000 disengagements for the same reason is different than 100 disengagements 10 times each.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Autopilot 2.0 is dangerous
- Thread starter Tribaltech
- Start date
I have driven long trips with my AP1 S, and am now completing a 6,000 mile trip in an S with AP 2. There is no question that AP 2 is well below AP 1 for
autosteer. In snow or rain the TACC of AP 2 is far superior to that of AP 1.
Across Wyoming on I 80 with the usual high winds, plus heavy rains and some snow the TACC performed magnificently. It slowed for winding roads more than I would but not too much to tolerate. In the same conditions Autosteer worked well on straight stretches even with rain.
However, AP 2 autoseer cannot ever be trusted to do anything correctly. It will never consistently do lane changes and spontaneously tries to go off the road without apparent cause.
It is long past the time to make AP 2 become at least the equal of AP 1.
This is my only real complaint with the newer cars. Otherwise the new ones are better than my P85D. My 6,000 miles in the last three weeks with the new one, a 75, make me want to get a 100D or even a P. AP 2 will not make me wait, but it certainly is not a positive factor influencing my thoughts.
autosteer. In snow or rain the TACC of AP 2 is far superior to that of AP 1.
Across Wyoming on I 80 with the usual high winds, plus heavy rains and some snow the TACC performed magnificently. It slowed for winding roads more than I would but not too much to tolerate. In the same conditions Autosteer worked well on straight stretches even with rain.
However, AP 2 autoseer cannot ever be trusted to do anything correctly. It will never consistently do lane changes and spontaneously tries to go off the road without apparent cause.
It is long past the time to make AP 2 become at least the equal of AP 1.
This is my only real complaint with the newer cars. Otherwise the new ones are better than my P85D. My 6,000 miles in the last three weeks with the new one, a 75, make me want to get a 100D or even a P. AP 2 will not make me wait, but it certainly is not a positive factor influencing my thoughts.
May I ask what firmware update you are using on AP2? I was surprised by the large variations in your experience! I don't doubt it, but surprised because AP2 has performed so well for me on road trips, it makes me wonder if something is wrong with your hardware or if the roads I was driving are clearly more traveled and have some better maps data (this explanation doesn't seem likely)... I will say that I can see AP2 failing miserably on some curvy mountain roads, but it is much much more reliable on highways for long stretches for me and does lane changes very well. I too will say that my car seems to be at it's best in Autopilot in rain and at night. Definitely has some inconsistencies with demanding curves and speed limit changes.... also two lane roads are definitely not reliable. I do wonder if that is what I'm missing with the AP2 to AP1 comparison.... I drive primarily on divided highways, and I don't trust the car on two lane roads... I wonder if AP1 is much better in this way... I really can't fathom AP1 being better than AP2 on regular highway driving, but I'm sure there are plenty who disagree.
Well @buttershrimp I finally have had some AP2 experiences in the last couple days on divided highways driving from Madison to Chicago that made me retract my claim that AP2 is on par with AP1. Several times during the trip south stretches of highway with fresh tar lines (like repair lines that looked freshly filled before winter comes) tripped it up severely - and stretches with poor lane markings did as well. About 70 miles out from Chicago the performance got markedly better - but so did road markings.
Oh and I'm on 17.34.something. It does seem like perhaps some reversion is happening. I trust these issues will be solved in a few months but in the mean time I've stopped trusting AP2.
I'm definitely getting past my unbridled euphoria towards AP2 now that I'm nearing 13,000 miles.
For instance, I'm beginning to feel frustrated with every Elon tweet related to SpaceX instead of autopilot... doesn't he realize my car needs rain sensing wipers and that Mars isn't going anywhere?
Sounds like I'm right on track for my Tesla maturation process.
By the way, I'm renting an AP1 car in North Carolina in a couple of weeks and am looking forward to comparing the hell out of it to AP2, it's been a while since I've done this.
Mark Z
Active Member
I drove the 2017 MS 90D back from service today. They found nothing wrong. They "test drove vehicle and could not duplicate customer concern. Auto pilot is working as designed. Camera pitch is within specifications and calibration is at 100%."
Next time it changes lanes by itself with visible lane lines, the fault lies with the software. My AP2 Tesla is perfect!
Next time it changes lanes by itself with visible lane lines, the fault lies with the software. My AP2 Tesla is perfect!
Last edited:
We've been running on .32 since before Harvey hit us (.36 finally installed this afternoon).
On a well market road, with lanes lines on both sides of the lane (i.e., not driving next to the exit/entry lanes), AP2 seems to work well. And is relatively "silky smooth" - without the jerking back and forth inside the lane that was present in earlier releases.
But, if the lane lines are only on one side, the lane lines aren't well marked, there are gaps with entry/exit ramps, or there is current/recent construction where there are temporary lines present - AP2 seems to struggle - and has tried to drive into traffic barriers several times.
While it's disappointing that we're now 9 months past when Tesla first promised to have AP2 running at AP1 level (by the end of December last year), they are making slow progress.
Though it's taking so long, if Tesla contacted us today to finalize the order for our Model 3, I'm not sure we'd be ready to pull the trigger and spend another $60K on a Tesla with AP2.x hardware, that isn't working yet with our Model S.
Since everyone knows or should know that AP2 software is ready for prime time yet, it would help considerably if Tesla was more transparent about the current state of the AP2 software. The release notes for each release should provide documentation on what has changed in each release, as far as AP2 is concerned - and the areas that remain as known problems that haven't been addressed.
Providing this additional information would help to inform drivers of the limitations of the current software, and help them to more closely monitor AP2 when operating under those conditions.
And, please give us a revised estimate on when Tesla believes AP2 will be at least as safe as AP1. If it's going to be another 6 to 12 months - while that's disappointing, I'd rather Tesla be honest about it - and properly set everyone's expectations - than for us to continue to guess what's changed in each release and how far we might be from getting AP2 to meet the goals of the EAP product many of us have already purchased...
On a well market road, with lanes lines on both sides of the lane (i.e., not driving next to the exit/entry lanes), AP2 seems to work well. And is relatively "silky smooth" - without the jerking back and forth inside the lane that was present in earlier releases.
But, if the lane lines are only on one side, the lane lines aren't well marked, there are gaps with entry/exit ramps, or there is current/recent construction where there are temporary lines present - AP2 seems to struggle - and has tried to drive into traffic barriers several times.
While it's disappointing that we're now 9 months past when Tesla first promised to have AP2 running at AP1 level (by the end of December last year), they are making slow progress.
Though it's taking so long, if Tesla contacted us today to finalize the order for our Model 3, I'm not sure we'd be ready to pull the trigger and spend another $60K on a Tesla with AP2.x hardware, that isn't working yet with our Model S.
Since everyone knows or should know that AP2 software is ready for prime time yet, it would help considerably if Tesla was more transparent about the current state of the AP2 software. The release notes for each release should provide documentation on what has changed in each release, as far as AP2 is concerned - and the areas that remain as known problems that haven't been addressed.
Providing this additional information would help to inform drivers of the limitations of the current software, and help them to more closely monitor AP2 when operating under those conditions.
And, please give us a revised estimate on when Tesla believes AP2 will be at least as safe as AP1. If it's going to be another 6 to 12 months - while that's disappointing, I'd rather Tesla be honest about it - and properly set everyone's expectations - than for us to continue to guess what's changed in each release and how far we might be from getting AP2 to meet the goals of the EAP product many of us have already purchased...
Interesting, I found that 17.34 was very very good at not trying to exit lanes on the highway with one lane marking... the latest update to 17.36 has struggled with this again... two steps forward, one step back is the pace I'm noticing. 17.36 seems to have had a marked improvement tracking through intersections at lights and in tight spaces on narrow roads moving with traffic in adjacent lanes. It doesn't search as much either when cresting over a hill, but is not there yet... I've had a couple of clearly confused moments when the car is seeing phantom lane markings, but I almost always can see why... for instance, recently repainted lane markings in a slightly different location, etc.
boonedocks
MS LR Blk/Blk 19”
I'm definitely getting past my unbridled euphoria towards AP2 now that I'm nearing 13,000 miles.
For instance, I'm beginning to feel frustrated with every Elon tweet related to SpaceX instead of autopilot... doesn't he realize my car needs rain sensing wipers and that Mars isn't going anywhere?
Sounds like I'm right on track for my Tesla maturation process.
By the way, I'm renting an AP1 car in North Carolina in a couple of weeks and am looking forward to comparing the hell out of it to AP2, it's been a while since I've done this.
Many of us that have had Tesla's for a while and at on our 2nd and 3rd are glad Elon stopped his "Tweeting" about AP. It did nothing more than lead to frustration and absolute lack of belief in anything coming from Elon. I for one publicly asked on several forums as well as emails to my SC and Tesla corp for Elon to stop his misinformation and (I won't call them lies) but untruthful expectations. The crickets chirping from Elon's twitter account are much better than the, next week, 3-6 weeks, this weekend, 3-6 months, SilkySmooth, SomethingSpecial....and on and on and on. I do hope that something major is coming along soon and that the likes of GM touting a coast to coast SuperCruise demonstration will be a kick in the pants.
Actually the promise (expressed as an‘expectation’) wasn't just for parity with AP1, it was for the whole array of EAP functionality: use of four cameras, automatic transition from one freeway to the next, etc. We shouldn’t forget the depth of the deception. That all sounds ridiculous now that we’re still struggling to match the old technology.
How crazy does that self-driving Tesla video appear now? Lol. We have the first cars known to man that will be self-driving BEFORE the auto sensing rain wipers are released.
Since we're not hearing anything from Tesla - either they are close to a major release, and waiting to make a big splash on that release.
Or (what may be more likely), progress is even slower than they expected, and they don't have anything new to say right now...
I don't fault Tesla for the demonstration video. For major software projects, it's not unusual to do a proof-of-concept demonstration, under controlled conditions with only a small amount of the software implemented. Though Tesla could have provided a better explanation on what the video is actually showing - and better set expectations on how much work Tesla had left to do before they'd have AP2 operating as expected.
Or (what may be more likely), progress is even slower than they expected, and they don't have anything new to say right now...
I don't fault Tesla for the demonstration video. For major software projects, it's not unusual to do a proof-of-concept demonstration, under controlled conditions with only a small amount of the software implemented. Though Tesla could have provided a better explanation on what the video is actually showing - and better set expectations on how much work Tesla had left to do before they'd have AP2 operating as expected.
Tjhappel
Member
It's getting very very annoying waiting for a decently reliable update. The biggest issue I have with AP2 is the lack of consistency day to day. I more or less drive the same routes to work on a day to day basis amd AP2 will handle it completely different sometimes..... not all the time but enough to not trust it so it's not worth turning on because it's not really a drivers aid at that point, just a gimmick. Ap1 was dependable for me, it made the exact same mistakes in the exact same spot so I always knew where and when to hold the wheel. AP2 is full crap shoot.
It's not getting better and now AP 2.5 is out. Elon should comment on the hold up to parity at least and then stop saying anything about AP.... I love my car but I am super mad at myself for shelling out money on fsd that's vapor ware.
It's not getting better and now AP 2.5 is out. Elon should comment on the hold up to parity at least and then stop saying anything about AP.... I love my car but I am super mad at myself for shelling out money on fsd that's vapor ware.
Electricfan
Active Member
You should be mad at Tesla (heresy, I know, to say that on this forum), not yourself. I'm mad at myself, and shouldn't be, for exactly the same reason - I upgraded my 2013 S for a 2015 as soon as I read about AP. Now I have it and its only good for long-distance trips. My commute to work (35 miles mostly on divided highways in Houston), the whole reason I took the financial hit on the trade, is beyond AP1's capability. At least I didn't shell out when I saw AP2.0 - almost did, but barely contained myself.
I'm buying the first car that comes out with FSD - that really works and can be test-driven, be it electric or gas powered, American or foreign made. FSD is it, for me. If you talked to anybody who knows me they would tell you I've been saying humans should not drive cars all my life. Its a basic belief of mine. I'm glad car-makers are finally making a serious effort to get there, and I credit Tesla for spurring the movement. But not for deceiving me about what AP would do when they were advertising it on their website in 2014/2015.
Forget full self driving for now. If my car would follow my inputted route to the GPS reasonably well while I had my hands on the wheel I'd be happy. Even if it couldn't make all the correct detailed decisions along the way but followed the selected route, exited in all the proper places, and made all the correct route decisions of new directions, then bravo.
That should be easier to accomplish and a god starting point that could please a lot of people. Now throw in TACC, some blind spot protection and collision avoidance, and a few other simple driving aids and I think people would be more content. Baby steps. Guardian angel mode and semi autonomous route guidance for inputted routes and you shut me up for a year.
That should be easier to accomplish and a god starting point that could please a lot of people. Now throw in TACC, some blind spot protection and collision avoidance, and a few other simple driving aids and I think people would be more content. Baby steps. Guardian angel mode and semi autonomous route guidance for inputted routes and you shut me up for a year.
Last night the 75 I was driving went back to it's owner, after 6,010 of my miles during the last three weeks. During the last two days,I drove about 800 miles from Kansas City to Little Rock to Atlanta. I no longer have access to the car and I don't recall the software version, although there was an update about two weeks ago.
The last two days, with no new update, had AP experience very different than it had been. On the somewhat curvy, US 65 from Springfield, MO to Little Rock, the AP handled even construction areas quite smoothly to the extent that it discerned proper lanes on newly paved asphalt which had the Botts dots installed but no other lane markings! Then on the construction-strewn I-20 between Tupelo and Atlanta it even managed most of the diversions, pretty well marked but with frequent pavement changes and asphalt/concrete transition. On this, my last hours with the car, it acted like a petulant child, remorseful, so behaving perfectly.
I have no way to explain this change. No new software, but the weather was hot and clear while most of the rest of the trip was cold with Pacific Northwest smoke then wind and rain through much of Utah, Wyoming and Kansas. Road markings were excellent through most of that area.
By contrast, on most of 1 and 101 in California and Oregon I did not even try to use it. On Canada 99 Vancouver-Sequim I could not even use TACC because it slowed and speeded erratically on curves.
As I try to understand the recent improvement without metaphysics I end out with wide lanes, mostly excellent pavement and markings, with clear, hot weather.
Were my experience with AP 2 to be limited to the last 800 miles or so I would be quite complimentary. It could not make consistent lane changes and was inconsistent with speed limits but those were the only issues. It even acted reasonable when cresting hills!
It's only fair to describe the positives along with the negatives.
how are you having problems on a divided highway in commute traffic?
for most people that's where AP (1 and 2) works great.
NerdUno
Member
Electricfan
Active Member
In traffic jams - yes - its worth it weight in gold. But I leave early so I don't get in heavy traffic. On any route I take there are points where AP1 cannot handle the road at 65 mph or higher - it swerves into the adjacent lane and if I didn't grab the wheel I'd side-swipe somebody. Randomly it will also just lose its way, not very often but too often to trust it.
But on long straight roads between cities it works great. I drive to Florida and TN from Houston occasionally, and it really makes the trip easier. I wish that was worth the price I paid, but those trips are way too infrequent for that to be the case.
Similar threads
