I'm hoping once the two stacks are merged, that will improve.
The other problem I have on high ways is I can't really use it in stop and go traffic. If I do, it will frequently accelerate to fast from being stopped then have to break hard when traffic reaches a stop again. Too jerky for my comfort.
Overall, when I can use it, it makes driving less stressful. I drive 600 mile round trips a few times per year and am able to have it engaged vast majority of the time on highways.
FSD Beta on city streets requires too frequent overrides for it to be anything more than a party trick at the moment. Even in very rural areas it will have problems of either not changing lanes when it should or changing lanes when it shouldn't.
Auto parking worked great for me initially but something changed and it is never able to detect parking spots for me. The few times it does it just flickers on the screen then disappears, resulting in me not being able to use the feature. I'm not sure if this is because of a problem with sensors or if it's related to me being in the FSD Beta. I've been in the FSD Beta for over a year now.
110% this.
In addition to this, I often find that it is a bit timid on the acceleration once traffic starts really moving. Like, it'll jerk a bit forward from a stop, but then veeerrryyy slllooowwwlllyy speed up, even once traffic ahead is over 100 feet away.
Of course it's not 1:1. AP has other features, such as a much nicer UI that better reports what it thinks it sees, and I think has sexy lane changing features and such.
But it makes me feel that FSD was more business necessary to keep a competitive advantage against cheap cars. Soon many cheap cars will have AP-like features at comparable fidelities.
If I can't trust AP with simple road rules, then why bother upgrading to FSD?
This suggests a trivia question, which I do not know the answer to: what lane in the US goes the farthest without a split or dead-end? Similar question for other countries.
I'd count most freeway exit ramps as splits from the adjacent lane, but would not count intersections with cross streets as splits unless the two roads intersect at a small angle. It is the actual lane that matters, not its name, so if a lane has a name change when it crosses some political boundary it still counts as the same line.
My first guess would be it is on some long freeway like I-10. My second guess would be it is on some rural road through the middle of nowhere.
I-80 starts in San Francisco and ends in New York City, but I'm certain there are splits that would make you change to some other highway if you stayed in the left lane.
Likewise, I-90 starts in Seattle and ends in Boston and travels through Montana.
My guess would be I-90, just because of Montana.
> Comparing Telsa's disengagements for its FSD beta, which can drive anywhere, against autopilots that work only on (certain) highways is not very comparable. At least compare highway to highway disengagements.
I own a Tesla, but do not have FSD (Not interested in it at all beyond being a party trick). After 3 years and 20,000 miles, I can say that I've never had to disengage AP because of imminent danger, though I HAVE had many times where the lane is splitting and AP can't decide which way to go, so I had to take control. Likewise, when a lane is merging, it would sometimes jerk left and right a bit and I'd take over. Both cases were more for comfort than safety.
Total, I'd probably average 5 miles per disengagement, but that's just so I can change lanes since the base AP does not include lane changing.