Autopilot observations from road trip

I just returned from a 4,900-mile road trip from FL to New England. I used Navigate on Autopilot (NoA) for most of the highway driving and Autosteer/TACC for the rest. I'm not going to write a report of the trip, but I do have a few observations about my AP experience.

My car is a late-2018 Model S100D with FSD and HW3 (upgraded to HW3 in February). The firmware was 2020.36.3.1 for the first part of the trip and 2020.40.3 for the last part. I didn't notice any difference in AP behavior between the two versions.

The most interesting observation was about Auto Lane Change. I've seen reports in this forum about Auto Lane Change either not working at all, or aborting after starting to change lanes, both with and without NoA. Auto Lane Change will only work if the car detects the outside lane marker of the target lane, and I found that when the outside lane marker is yellow, which the leftmost lane marker is on much of I-95, the car would often have trouble detecting it. Whenever this happened, I had to manually override AP to change lanes. This also impacts how NoA proposes lane changes, because it won't propose one if doesn't see that outside lane marker.

NoA proposed lane changes are still hit and miss for me. For that reason, I've never disabled lane change confirmations. Sometimes NoA will not propose moving into the passing lane to get around a slower vehicle even when the outside lane marker is detected (which you can tell from the visualization). Other times it will propose moving into the passing lane when you're pretty far behind the vehicle ahead of you and you're not closing quickly. Sometimes NoA won't propose moving out of the passing lane after passing a vehicle (though it does so more reliably when there's another vehicle behind you). Sometimes NoA will want you to pass on the right, which isn't always legal.

On several occasions, NoA wanted me to change lanes in order to follow the route, but no lane change was necessary. I'm not sure if this was always related, but I noticed that when there were shadows across an adjacent lane, the visualization sometimes displayed "phantom" merge arrows in the lane, the same kind that used on a lane that's about to merge into another.

There were a couple of occasions where the car detected traffic lights on the interstate where there weren't any. In one case I was crossing the Tappan Zee Bridge (aka Governor Mario M. Cuomo Bridge) and the green arrows showing which lanes were open were misinterpreted as traffic lights. In another, I was following a truck that had red taillights on the upper part of the cab and my car saw those as red traffic lights. Fortunately, the car didn't try to react to the "traffic" lights in either case.

There's always been a lot of discussion about phantom braking. That happened to me several times on this trip, once caused by a small bridge/overpass, the rest by my car thinking a car in an adjacent lane was moving into my lane. Nothing unexpected, and the car never "slammed' on the brakes as others have reported. I occasionally experienced the "feature" where AP slows down if you pass slower moving cars, but it was infrequent.

Although AP is supposed to slow the car down in sharp curves, it doesn't always do that, so be careful. NoA is really bad about slowing down for exit ramps that curve (it doesn't always pay attention to exit ramp speed limits), so although I tried using it many times, I generally disabled AP before taking an exit because I didn't trust NoA to handle it.

I didn't see where the new ability to read speed limit signs made much of a difference. Sometimes the speed limit in the visualization was wrong, but I couldn't tell if that was caused by bad sign reading or bad GPS data.

Something I ran into that will present a challenge for FSD: part of I-95 S was closed for construction, but the navigation system didn't know that, even though Google maps did. I followed detour signs on back roads to get back on I-95, but the car kept trying to route me back to I-95 and the next available opportunity, even though those onramps were closed.

All in all, AP worked well for me on the trip, and it made driving a lot less stressful. I'm always vigilant when using AP, so I'm not too worried about it messing up since I can intervene quickly. I tend to give AP the benefit of the doubt even in dicey situations (e.g., construction zones), and only intervene when necessary. My first couple of FL to New England road trips were with an older S with AP1, and even that worked pretty well, though it had a lot more trouble following lane markers in construction zones and other places where the lane markers weren't very visible.

I don't know when true FSD will be here, but for me AP on the highway is much more important, and I'm pretty satisfied with how it works today, and it only gets better with time.


  • > @EVRider said: Nothing unexpected, and the car never "slammed' on the brakes as others have reported."

    It's the Single Pedal Regen that kicks in which is solid brake that is unexpected for driver and all drivers around the car who also suddenly react to car braking for no reason at high speed.
  • I just posted a couple of things with FSD I noticed including not slowing down enough for exit ramps and some lane change issues.

    As you stated it’s still not really full self drive but a tool that works most of the time but you need to stay aware of everything around you in case it does get confused
  • @EVRider - thanks for the report. I agree with you that the system works as designed and I getting better all the time.
  • getting better....
Sign In or Register to comment.