How does the new 10.3.1 FSD beta profile (chill, average, assertive) interact with the previous (and still exists) AP following distance setting (setting 1 thru 7 using the right scroll wheel) and speed-based lane changes setting (mild, average, Mad Max)? Does the new FSD beta profile override those other two, apply sometimes while the other two apply other times (I can't think of when one would be used vs the other), or somehow merge/average these settings together? At least for exiting the passing lane, I can see how that setting is now only applied if FSD beta profile is either chill or average (and assertive would override whatever exit passing lane setting you have).
In any case, was thankfully able to download and install 10.3.1 before heading out yesterday for more than a total of 125 miles. Some highway, a lot of dense urban city, and even newly constructed roads and unmapped areas. A few things I've noticed since 10.2 and compared to before:
1) I got phantom braking pre-FSD beta only once in awhile. Shadows, overpasses, etc. rarely seemed to phase my radar-equipped 2018 Model 3. It did happen, but infrequent enough that my guess is maybe every couple months I'd experience it? It was worse when I first got the car, but seemed to be less frequent over time. Granted, this car has only driving in California so maybe it's "tuned" to our tree shadows and overpasses better?
Ever since 10.2 FSD beta (when my car supposedly switched to vision-only), I haven't experienced any hard phantom braking at all. I have once in awhile experienced a slight slow down. It's gentle, and if no one's behind I let it continue so I can try to understand what it's trying, but sometimes I end up overriding it with the accelerator pedal (not fully disengaging FSD beta/NoA), then letting up a few seconds later and it takes over/resumes just fine. Strange.
2) I drove up some VERY steep neighborhood roads that have a 4 way stop at the top. This is the only time I've experienced it stopping WAY too early. I was out of my seat trying to look over the car hood to monitor what it was going to do. It seemed to treat the stopping line way too early as well, meaning once it stopped and thought it was clear (I'm not confident it could see if I couldn't), it accelerated through the 4 way stop as if it was our turn. There was once that pedestrians were crossing in front, which the car seemed to hesitate for, but the instant I felt it accelerating even a little as if it was safe to go through, I hit the brakes to disengage because it was maybe not gonna hit them, but too close for comfort. (And I reported that as well.)
3) It does well in the rain. There was one instance of maybe a couple minutes it complained about poor weather, but the warning went away and NoA came back just fine.
4) Vision only NoA on highways seems fine. I wish 80 wasn't the limit but can understand the caution at this time.
5) It was handling most stop signs & right turns on red much better than 10.2, but there were a couple it creeped so slowly again that I ended up taking over to not spook the people waiting/behind me.
6) In an industrial warehouse area with hardly any cars but a lot of new construction (including roads closed and new roads created), I did a demo for some clients that were curious. It went very poorly.
The navigation was way off because of all the road changes, and that just messed with the FSD beta like no other. It was interesting and kind of amazing to see when FSD beta's realtime vision and path planning would override the navigation and miss a turn, forcing the navigation to reroute, but all too often it was just so confused and cautious, and sometimes decide to follow the navigation which meant barreling towards a newly placed cement curb, barrier, etc. and I'd have to override as my passengers screamed. Needless to say, all my hyping of how reliable it was most of the time and only needing few disengagements was somewhat discredited in their eyes.