TMC is an independent, primarily volunteer organization that relies on ad revenue to cover its operating costs. Please consider whitelisting TMC on your ad blocker and becoming a Supporting Member. For more info: Support TMC
Start a Discussionhttps://teslamotorsclub.com/tmc/tags/

AP 1.0 vs AP 2.0 Functionality Tracking

Discussion in 'Autonomous Vehicles' started by napabill, Aug 30, 2017.

  1. napabill

    napabill Active Member

    Joined:
    Apr 30, 2009
    Messages:
    1,389
    Location:
    Napa, California, United States
    Is there a thread anywhere that tracks the relative functionality currently in place for each hardware suite? I currently have AP 1.0 and will be looking to upgrade to a new MX, but am concerned that the AP 2.0 functionally is still trailing behind AP 1.0. And of course, now we have AP 2.5 to consider.
     
    • Like x 2
  2. mrkisskiss

    mrkisskiss Member

    Joined:
    Jan 26, 2017
    Messages:
    143
    Location:
    London
    It's such a moving target that any such thread would be out of date within two weeks. There is a thread on the Tesla forums that attempts to track AP2 progress - I agree that a similar thread would be good here, but it's hard to get definitive answers when there's so many firmware versions flying around and the minor improvements to auto-steer are so subjective and context dependent.

    As of 2017.32.6, the differences from AP1 are:

    - No auto/rain-sensing windscreen wipers
    - All objects are displayed as cars on the dash/IC - we don't have separate graphics for trucks, motorbikes, pedestrians or cyclists although the system does appear to "see" them (just not represent them on the dash). They are only displayed in a single lane, except during lane changes (as opposed to seeing representations of the cars around you in all lanes on highways with AP1). There is no evidence to suggest that the car doesn't see them - it just doesn't display them to you on the dash.
    - Speed limit signs are not read by the cameras, instead they are pulled from the map data.

    Functionally speaking, AP2 is now the same as AP1 in terms of actual driving features. As of this FW version (17.32.6), AP1 is still slightly smoother and reliable in all situations, although it's very close now - especially on the highway. Phantom braking (a sudden deceleration underneath some bridges, overhead signs etc) seems worse on AP2 - most people have experienced it at some point - but it can happen on AP1 as well. It is improving.

    Bottom line: unless you're bothered about seeing trucks/bikes on the dash and having rain sensing wipers, there's not a lot in it these days. I wouldn't let it put you off upgrading your MX.

    If you order a new MX today, you'll probably get the new 2.5 radar and redundancy processors etc. What difference this makes in practice is almost impossible to tell at this stage (presumably not much)
     
    • Informative x 1
  3. napabill

    napabill Active Member

    Joined:
    Apr 30, 2009
    Messages:
    1,389
    Location:
    Napa, California, United States
    Appreciate your reply and was exactly what I was looking for. My MX lease still has a year+ to go, so won't be needing to make any decisions for a while. Hopefully our M3 will arrive towards the end of the year so will have some hand's on experience with that.
     
  4. alcibiades

    alcibiades Member

    Joined:
    Apr 27, 2017
    Messages:
    471
    Location:
    IL
    My advice -- for what it is worth, hey-o! -- is to spend time with an EAP car personally. The reports from owners differ wildly. My own experiences have been dreadful.
     
  5. AnxietyRanger

    AnxietyRanger Well-Known Member

    Joined:
    Aug 22, 2014
    Messages:
    8,022
    Location:
    EU
    I believe a lot of people would disagree with that.

    The level of confidence AP1 instills in you is significantly higher than that of AP2. Ghost braking is just one of the things, following phantom lines for example is another. With AP2 you simply have to keep paying attention a lot more at this time.
     
    • Like x 2
  6. mrkisskiss

    mrkisskiss Member

    Joined:
    Jan 26, 2017
    Messages:
    143
    Location:
    London
    Oh, I definitely agree with you - AP1 gives you a different, more confident feeling... but AP2 (in my experience) is no longer bad enough to purposefully avoid upgrading the rest of the car at this stage (in my opinion, of course!)
     
  7. boonedocks

    boonedocks Member

    Joined:
    May 1, 2015
    Messages:
    809
    Location:
    Gainesville, GA
    +1AnxietyRanger. I never get the nags or hold wheel with my AP2 because for fear of it misbehaving I hold the wheel tighter than when I am driving. Still a long way to go....bye I am officially 5 FW revs behind now. Even after calling Tesla to inquire why AND having open investigations into AP2 / TACC issues. Kinda throws a WDGAF vibe from Tesla right now. ¯\_(ツ)_/¯
     
    • Like x 1
  8. jlarmstr

    jlarmstr Member

    Joined:
    Jun 28, 2017
    Messages:
    84
    Location:
    Portland, OR
    I have only driven an AP1 car for a short while (loaner during a service), I do think AP2 is still behind AP1. However, I started driving AP2 cars shortly after they were launched, and the rapidness of the improvements in stability is very apparent. I would venture to say AP2 is within 80-90% of that of AP1 at the moment (i.e. following distance, smoothness of lane changes, coming to stop in stop and go traffic and so forth), totally subjective assessment so don't read too much into it. However, it is clear that AP2 is behind AP1.
     
    • Like x 1
  9. napabill

    napabill Active Member

    Joined:
    Apr 30, 2009
    Messages:
    1,389
    Location:
    Napa, California, United States
    Thanks for the report. Now all you have to do it repeat monthly for the next 12 months!;)
     
  10. tescroft

    tescroft In Musk, we trust.

    Joined:
    Aug 23, 2015
    Messages:
    124
    Location:
    Switzerland
    AP2 is in production cars for almost a year now. Tesla is really acting borderline here by not delivering features like rain sensoring for such a long time. First, I was pissed getting an AP1 car in september 2016, right before the AP2 annoucement/production launch. Now I've to say I am happy that I got an AP1 vehicle, as the system is still superior to AP2 one year later and it does not seem that AP2 will significantly surpass AP1 in my personal ownership of the car. While it is a wise desicion to kick out Mobileye from a strategic point of view, it is not for customers in the short term.

    For me, a reason to switch to AP2 or AP2.5 would be when AP2 is able to drive safely and speed-adjusted through freeway construction work sections hands-free. That would enable me to drive almost my whole commute hands-free. In my impression, the AP2 production software is still quite some time away to be able to do level 4 highway driving.
     
    • Helpful x 1
  11. BLKTSLA

    BLKTSLA Member

    Joined:
    Sep 22, 2016
    Messages:
    118
    Location:
    Unknown
    I think AP2 is still behind, not in terms of potential/capability, but in terms of what is actively being used. I did a side-by-side comparison here:

     
    • Like x 3
  12. croman

    croman Active Member

    Joined:
    Nov 21, 2016
    Messages:
    1,764
    Location:
    Chicago, IL
    Eventually Tesla will start rolling out the E and FSD parts of the $8k packages they sold. Then AP2 will be truly superior but right now AP1 is still better in most ways. AP2 does a better job for me on unmarked roads, though that's debatable as AP2 has regressed considerably since 17.17.4 in this regard. Mostly AP2 does the job just fine. I agree with the earlier 90% assessment. Its almost at parity but that was promised back in December 2016. Keep this in mind to anyone upgrading to a new hardware suite by Tesla -- you will almost assuredly not get what you think or what Tesla is saying you should. They'll say anything to get your $$ and once they have it, you're at their mercy in terms of a timeline or whether it even happens at all.
     
  13. AnxietyRanger

    AnxietyRanger Well-Known Member

    Joined:
    Aug 22, 2014
    Messages:
    8,022
    Location:
    EU
    #13 AnxietyRanger, Oct 16, 2017
    Last edited: Oct 16, 2017
    So, @Tam and @Pale_Rider, since we've discussed AP2 steering and nags in previous threads I thought I'd post an update.

    Over the weekend I took my Model X (2017.36, latest available for my car) on a highway that has a Supercharger alongside it. I thought maybe this stretch of road would be better "learned", if there indeed is such learning going on. I assume there is some learning at least on overpasses and ghostbraking...

    As you recall, AP2 for me is an unreliable piece of crap on the local stretches of highway and some surface streets I usually test it on. For example, I've had divided-highway incidents of AP2 following tar lines towards concrete walls (ignoring clear lane markings) and following (white-marked) lead cars to adjacent lanes and then disengaging...

    I already reported that @Pale_Rider's suggestion to keep just one hand on the wheel helps with getting rid of the nags - and mostly it does. Not 100%, but I would say a massive improvement - to the tune of 95% perhaps.

    Turns out, it was better steering-wise too and pretty much in every way. EAP was actually pretty nice on this road.

    There is no difference in road quality, markings or the like compared to my local roads. If anything, weather was worse than I usually test on. But I guess it must be a more traversed road by Teslas, because I had no ghost brakings from overpasses (only a few from cars in adjacent lanes in curves) and the steering was solid almost all of the time.

    Or maybe there was just some other, unknown quality to the road or conditions that was better for EAP...

    If THAT trip was my only expeirence with AP2, I might join the choir of saying people are just making the negative experiences up...

    Some other notes: I saw cars on adjacent lanes in the IC when changing lanes, but not otherwise. I did not see any blue lead cars. :) I did not see any separation by car type on the IC and definitely no recognition of traffic signs, simply an outdated database.
     
    • Like x 2
  14. AnxietyRanger

    AnxietyRanger Well-Known Member

    Joined:
    Aug 22, 2014
    Messages:
    8,022
    Location:
    EU
    After my recent experiences and readings on TMC, I am happy to say I don't think the different experiences on AP2 are solely down to bias. They are also down to where one has been driving it and what exact software versions one's car has received. IMO that explains a lot of the different reactions to AP2. Of course people naturally (knowingly or not) filter or are selective by bias too, but I genuinely feel this system is working differently in different locations, I can't see any other explanation to it.
     
  15. Pale_Rider

    Pale_Rider Member

    Joined:
    Jul 28, 2016
    Messages:
    158
    Location:
    Houston, TX
    In my little area of the world, AP2 pretty much performs exactly like AP1 does (aside from astetic things like showing vehicles in multiple lanes) except in on annoying area, where it allows me to use auto-lane change. For some reason AP2 is not aware that there are multiple lanes (and I assumed this was based on map info in AP1) on many of the roads I drive on that AP1 was capable of. I’ve not spoken to the service folks yet to figure out why. Very odd though. Anyway in about a month I’m going to be driving from Houston to Northern Virginia, through places few Tesla travel, so I’ll keep an eye out for poorly behaving AP2 and see if I notice any correlation between driving performance and different areas. I will say I’ve already tried it on quite a few “back roads” in West/Central Texas and New Mexico and was surprised it was able to handle some things that AP1 wouldn’t not have handled. It is by no means perfect and has a penchant for occasionally trying to follow the tar lines on local roads near my house right into oncoming traffic at intersections but I can say AP2 performs to my expectations (good and bad) once I figured out its strengths and weaknesses...
     
    • Informative x 1
  16. jdjeff88

    jdjeff88 Member

    Joined:
    Aug 25, 2016
    Messages:
    125
    Location:
    Texas
    Only 4 days with ap2 (2.5) after 1 year with AP1 but so far it performs better in all of the challenging parts that confused the AP1 car. Driving Houston to Galveston with construction and temporary lanes in particular ap2 did great where I constantly had to disengage AP1. On 2017.40.
     
    • Informative x 2
    • Like x 2
  17. Pale_Rider

    Pale_Rider Member

    Joined:
    Jul 28, 2016
    Messages:
    158
    Location:
    Houston, TX
    We are in the same neck of the woods and you seem to be seeing what I’m seeing.
     
  18. AnxietyRanger

    AnxietyRanger Well-Known Member

    Joined:
    Aug 22, 2014
    Messages:
    8,022
    Location:
    EU
    I look forward to seeing what, if anything, 2017.40 does on this...

    That said, it does take more than 4 days with AP2 to see what's what. :)
     
  19. jimmy_d

    jimmy_d Member

    Joined:
    Jan 26, 2016
    Messages:
    190
    Location:
    San Francisco, CA
    I just did the upgrade about a month ago and am interested to compare notes. I drive mostly LA to SF, had about 25k miles on AP1 over 2 years and now have about 1500 on AP2. Most recent 100 odd AP2 miles on 2017.40 and the rest on 2017.38. I also have a HW2.5 car.

    Here's my sense of it below. If you have a moment maybe you could tell me where your experience has been different.

    My overall sense is AP2 is not as trustworthy as AP1 was as of a month ago, but it has some places where it is better. I'm especially pleased with accelerator control under AP2 - it seems to be a lot smoother at maintaining the space to the lead car in traffic with fewer sudden acceleration and braking events compared to AP1. AP2's management of lane merging events is really nice - it seems much better than AP1 in those situations. It almost never brakes unnecessarily whereas that was common enough on AP1 that I didn't really let it handle merges.

    The biggest single weak point I see on AP2 seems to be holding the lane center on curves, especially the second bend on an S curve. AP1 would move around some but it was really rare to see it cross a lane marking whereas I'm seeing AP2 do that frequently whenever the freeway gets more than slightly twisty. It also seems like AP1, at least as of a few months ago, was really good about maintaining spacing with vehicles in adjacent lanes but AP2 has me watching those gaps much more closely. It had probably been five thousand miles since I disengaged AP1 because it was moving too close to another vehicle but I'm seeing that happen with AP2 much more frequently - maybe every 30 miles in heavy traffic.

    Stop and go traffic with speeds staying below maybe 30mph (really common here in LA) are really good in both 1 and 2, but the improved accelerator control in 2 makes me a lot more comfortable with it. At low speeds AP2 seems to detect unmarked pavement boundaries like curbs and shoulders which AP1 didn't seem to be able to do. That mostly isn't helping much but I find it really interesting because it bugged me that AP1 was so lame in that area. I also notice that AP1 was really picky about what it would consider a 'lane' in terms of the spacing between lane markings. When you had lane merges or splits it would be more likely to get lost but then it seemed less likely to actually do something stupid - seems like it could coast through the uncertainty better.

    So far I haven't seen as much ghost braking for bridges on AP2 as on AP1, but I do get unnecessary braking when forward cars on adjacent lanes intrude on my lane. All in all it's a lot more ghost braking when I drive in the city and a bit less when I drive on rural limited access highways.

    Occasionally I will play with it on surface streets to see what it can do - keeping it on a short leash of course. Though both AP1 and AP2 seem woefully inadequate for surface streets where I live it seems like AP2 is overall making fewer mistakes, but when it makes a mistake it's more likely to be a scary one.

    In general that's probably a pretty good overview actually: curves on AP2 seem a lot worse (I mean, AP1 has gotten really, really good with curves this last several month) which is a big deal, but if you exclude that I think it makes fewer overall mistakes than AP1. And when AP2 makes mistakes they seem more consequential.

    Anyway - be interested if your experience is different.

    P.S. I approve of your color choice
     
    • Informative x 3
  20. jdjeff88

    jdjeff88 Member

    Joined:
    Aug 25, 2016
    Messages:
    125
    Location:
    Texas
    Thank you sir. You also have excellent taste. This was my first red car because I never wanted to be that guy with the red car. Now I am that guy...on his second red car.

    So far I have found AP2 to be as trustworthy or more. Not having had to go through the growing pains of the earlier releases maybe makes me less twitch as I essentially went from AP1 to AP2 .40 in 5 weeks. I know the areas and circumstances that gave me trouble in AP1 and without exception AP2 has handled them all. That includes a 15 mike stretch of HOV lane with multiple peel off points that gave AP1 fits and drove me nuts. The issues also included phantom breaking from overhead signs on hills. Didn’t have to disengage once with ap 2 and no phantom breaking.

    I have had two instances of brake checks when cars a bit ahead of me wandered into my lane. I didn’t see that as anomalous behavior as I was barreling along at 85 ish and the wander could have turned into a full on one incursion. The instance where I was paying most attention the car went from 86 to 81for several seconds and then resumed promptly as the car ahead veered back into its lane. Whole thing was pretty smooth.

    I have had good luck on surface streets as well. In the suburbs and downtown I haven’t had to disengage yet for the car deciding I needed to be somewhere other than where the nose was pointed but haven’t been to aggressive yet.

    I haven’t noticed any issues with curves yet. A minor correction or two on a couple but always within the lanes. And it does seem to handle lanes better overall, even when the lane markings are indistinct or disappear altogether.

    I do notice that the car seems more centered in the lane at all times, where my AP1 car was biased to the right in a way that sometimes made me uncomfortable when passing cars. It was my impression that to some degree this was causing the illusion of truck lust where in my normal car I seem to have biased myself more to the left side of the left lane and unconsciously shied away even more for trucks.

    Full disclosure though I only have 400 miles on my car as it has been in the shop for the last 4 days getting a list of things fixed. But I did take it on the tour of AP1 terrors promptly after I updated on the second day to .40 to see what would happen.
     
    • Informative x 2
    • Love x 1

Share This Page