Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Does 17.17.4 AP2 actually perform better for you?

Does 17.17.4 perform better for you

  • Yes

    Votes: 29 82.9%
  • No

    Votes: 6 17.1%

  • Total voters
    35
  • Poll closed .
This site may earn commission on affiliate links.

gearchruncher

Well-Known Member
Sep 20, 2016
5,120
7,153
Seattle, WA
Chris Lattner (head of AP at Tesla) tweeted "Not mentioned in the release notes, but performance/feel is much improved as well"

Does the latest release actually work better for you? It's still a mess for me on local roads.

Here's a 1 minute video of my first drive with it, from a dashcam placed right below the camera assembly in a Model X. It first dives for the outside of the road and over-corrects into the oncoming lane, then a few seconds later actually just goes into the oncoming lane, and then at the end it dives for the inside curb. All of this at 25 MPH and with clear lane lines.

It did all of this before, and does this repeatably, so I fail to see how this is improved.

Untitled - Streamable
 
It definitely performs better for me but it is still relatively weak on roads with broken lane lines (e.g. Local roads with intersections)

I would say the most dramatic improvements are for highway driving and tracking highway curves without departing the lane, plus nearly eliminating the drunken wandering effect on straight roads.
 
  • Like
Reactions: travis010
Interesting how different our experiences are. I never had issues with the previous version on the highway. It did a fine job, except that it darted for exits now and then and pointed me right at barriers. Never had an issue with lane positioning or handling highway curves. So I had assumed that any statement about it working better was in how it would handle non-highways.

Glad it is showing improvement for some.
 
Current version is a drastic improvement over previous (although still not perfect). From what I remember of an AP1 test drive, AP2 is getting pretty close, but not quite parity.

But I think the current version is now 'good enough' to demo to family and friends. They've asked before, but I thought AP2 was a little to 'squirrely' until now.

Highways: Improved (more refined), although last version was not bad. But last version had a sometimes-tendency to bounce back and forth slightly within the left/ride sides of the lane or do 'partial' corrections while in a curve. Now I noticed it's a much smoother curve.

Local roads: MUCH improved. I don't feel like I'm going to die at any moment. Given that, it's not quite as good or refined as highway, but it's much nicer (if there are good lines). Even a few missing lines around a slightly curvy and undulating up & down road works pretty well (at 35 mph).

Overall, I'm REALLY happy with this version... but also look forward to continued refinement (as there is still work to do).
 
Yeah if they could deal with lane lines disappearing and reappearing better, then it would be ideal for city driving. Right now if there's a curve and a break in the lane markings, it's totally unpredictable how AP2 will handle it. AP1 did a better job of following a reasonable path through the intersection.
 
AP1 way better. I am viscerally firghtened for my life using AP2 when in the left hand lane and there is oncoming traffic in the opposing lanes. Although alert with AP1, I'm not scared in that situation. I am sometimes driving AP1 and AP2 the same day as my wife's car is AP1 so I can really tell the difference.
 
Answers may vary due to how many revisions happened between updates, probably car dependent. Mine, for example, skipped 5 in-between updates (but not wifi related).

Testing 17.17.4 since yesterday.....can't vote yet

old |v8.1(17.11.3)| AP2 |
-
-
-
-
-
new |v8.1(17.17.4)| AP2 |
 
for those having wandering issues. I've noticed it was always when there was a seam in the road or in one case where there were black tire marks from 18 wheeler that had locked his brakes. I was wondering if lane wandering for others are a result of the road conditions. I am surprised that a small thin seam seems to make so much difference.
 
A European voice:

On motorways, there's an appreciable improvement. The tracking seems to be better, auto lane changes are faster and with less dithering and the lane changes with slower traffic in lanes one further to the right seem to be without the panicky braking where the AP didn't seem to know if the car/lorry in that lane was in the lane you were entering or not.

In towns, it's now finally usable in Europe, though still not 100% trustworthy! Vienna has tram lines all over the place which confused the hell out of earlier releases, but my drive to work this morning was much smoother without the car trying to kill me every few hundred metres! Nice change.

Still not where AP1 was on that first test drive last year, but it feels like a huge step forward, even if 17.17.4 doesn't come with auto full beams in Europe.
 
While mine is definitely better (no AP1 experience, fyi), I am wondering (after watching your video) if AP was sensing a side collision (because of the warning you got and the movement of the car - 17.17.4 was supposed to include side collision). Not that this makes it okay, but it seemed like it was not wandering because of lane misidentification, but instead sensing some danger/hazard. Have you driven the same road when conditions were different (it looks like trash day) to see if it had the same issues at the same places?
 
AP2 is progressing for sure, but as did AP1, its very incremental per release. Variability remains huge among AP2 drivers, conditions and descriptions....just read the threads. You won't find such deviation among AP1 drivers, although there was a time when it was similarly varied, comically so as we endeavored to find progress when there wasn't, or lay claim our favorite drive was finally nailed or our lingering frustration to 'get what we paid for' finally resolved. Heck we imagined tons of stuff and probably had engineers ROFL :D

AP2s leapfrog moment is approaching but it remains several iterations away for the finer tuning of basics to surpass AP1, while concurrently teasing us with early state features unique to AP2. I voted no because it gave some and it took some back.....which is the pathway to AP progress but it's not yet overall improved performance in my test lab. I logged a disappointing 1100+ miles in my AP2 17.17.4 over the past few days, and I can't wait to jump into my AP1 and relax today.
 
  • Like
Reactions: vandacca