EVNow
Well-Known Member
Not sure why they have 80 mph limit now - so, no idea what might fix that. But since there has been no talk of that in 11.x, I would not hold my breath.I hope we can get back to 90mph limits with FSD 11.x.
You can install our site as a web app on your iOS device by utilizing the Add to Home Screen feature in Safari. Please see this thread for more details on this.
Note: This feature may not be available in some browsers.
Not sure why they have 80 mph limit now - so, no idea what might fix that. But since there has been no talk of that in 11.x, I would not hold my breath.I hope we can get back to 90mph limits with FSD 11.x.
it's probably just a matter of collecting data of vision vs radar performance of AP at around ~80mph. Once vision consistently matches radar, i'm sure they'll open up the next step change. It was similar to when they did the 75->80mph unlock. It will probably happen randomly outside of any FSD related update. Could happen soon, could never happen...(hopefully not)Not sure why they have 80 mph limit now - so, no idea what might fix that. But since there has been no talk of that in 11.x, I would not hold my breath.
Who are you and what have you done with "optimistic, every update is mindblowing" powertoold10.10.2 seems to be a slight improvement over 10.10 but it's still doing stupid things that 10.8.1 didn't often do:
Chg lanes when there's a fast approach car behind
Randomly try to move out of lane without signalling when there's a fast approaching car behind
I don't think 10.10.2 would be safe to widely release. The lane chging behavior is accident prone.
Sadly I find that pretty much everything "auto" on my car is better turned off. These cars can do lots of things, just almost none of them well.But on highways you can turn off the automatic lane change behavior. Tesla should apply that UI setting to FSD beta also, at least until they can make lane changes on city streets work better.
You start to wonder if some of the SW notes are just lip service with these regressions. Appreciate the drive and updates on supposed progress. We’ll keep hoping.10.10.2 didn't make much difference in my brief testing today. In fact some things seemed worse:
* Freeway exit starts off as 1 lane, widens into 2, the car needs to stay left; it started off sticking to the left as before, but did a wild swerve to the right just as the road widened. Then it swerved back into the left lane. Has not done that before with the same exit, even before FSD beta.
* Left turn with 3 target lanes; leftmost lane is a left turn only, right 2 lanes are straight only. I needed to go straight. The car decided to turn into the left-only lane and stop behind the cars waiting for their arrow.
* 2 right turns around some pylons (tall white ones) it wanted to swerve into them so I disengaged. Previously it would get nervous around them but not try to approach them so closely.
* Couple left turns felt like they might clip the median so I disengaged. Both had a turn lane next to me with more cars; I didn't trust it.
* So much for right-of-way improvements - at 4-way stop it yielded to 2 parked cars (not unusual), quickly recognized that they weren't moving (improvement), but then continued sitting at the stop sign. I let it do so for 5-10 seconds until other cars appeared in the distance, then nudged it through. Haven't had the same issue at that intersection before, but maybe it's a fluke as I've had other intersections in my area sporadically fail the same way.
* Maybe the fault of new maps, but a freeway exit I've taken many times wasn't recognized, and instead of slowing down to 55 or 50 as normal, it kept going 75. After slowing down for some cars ahead it merged left and tried to accelerate back to 75. Finally dropped to 50 as it crossed the threshold where NoA would end in the past.
* Street in my neighborhood is now recognized as 30mph. It's a regular 25mph residential street with speed bumps.
The only true improvement on this drive was one unusual left turn that it has failed at a basically 100% rate; today it was smooth and the car pointed onto the correct street instead of aiming at either oncoming traffic or parked cars.
What did you have earlier ?10.10.2 didn't make much difference in my brief testing today. In fact some things seemed worse:
Those are what they found in simulation / testing. Any particular real word experience may be different. Usually some May see progressm some no difference and some regression. You just hope the former group is larger.You start to wonder if some of the SW notes are just lip service with these regressions. Appreciate the drive and updates on supposed progress. We’ll keep hoping.
10.8.1 which itself was not great in some ways, most of which were the excessive false cone detectionsWhat did you have earlier ?
10.8.1 had a few regressions for me which continued in 10.9. Will have to see what happens with 10.10.10.8.1 which itself was not great in some ways, most of which were the excessive false cone detections
This is what I don’t get. Am I misunderstanding something? It is my “assumption” (I know, I know) that there are supposed to be improvements albeit small iterations each time on all these areas. Or even SOME of these areas. Why is it that there are so many regressions in most if not ALL these areas with each update. It seems like there is either no progress or that we are moving backwards in most areas.I drove what has become my FSD torture test drive on 10.10.2 last night.
Seems like it has kept most of the bad things from 10.8.1 sadly
Approaching a T junction stop sign where the main road is curved. FSD ignores approaching cars from left and tries to turn right in front of them.
Next T junction where we have right of way, no yield, just a right turn. FSD still comes to a full and complete stop and edges forward before make an extremely slow turn. I usually take this corner at 20mph.
Next is the left turn at the 4way where we have right of way (no road signs). FSD comes to a complete stop again before trying to turn left into someones garden, then overcorrects and cuts the entire oncoming lane to get to the correct lane.
Then another left turn T with full right of way, another full stop and stupidly hesitant turn with lots of stops and wheel spinning.
All that in the space of 400 yards. Every single one completely useless.
On the drive back is was even worse. Literally zig zagging around the corner with wild steering wheel spinning.
I think this version is even worse than 10.8.1 on unmarked roads, driving right down the center and ignoring oncoming traffic.
I let it get to a left turn T junction, as soon as it spotted the road markings for the junction it freaked out an flashed the red steering wheel "Take over immediately".
That little 5 mile drive was 8-10 snapshot disengagements in each direction.
Hopefully will try other drives today, but not feeling good about 10.10.2 right now.
I've always assumed that local conditions such as weather, lighting, traffic and mapping inconsistencies tend to obscure most of the improvements from one minor version of FSD beta to the next. No matter how much I try to standardize my drives, the behavior of any version of FSD beta is not all that reproducible. Every so often a version contains a specific improvement that is noticeable, though. One example was the switch from friction braking to regen.This is what I don’t get. Am I misunderstanding something? It is my “assumption” (I know, I know) that there are supposed to be improvements albeit small iterations each time on all these areas. Or even SOME of these areas. Why is it that there are so many regressions in most if not ALL these areas with each update. It seems like there is either no progress or that we are moving backwards in most areas.
Ski
In fairness, no version of FSD has come close on this particular drive. Sadly for me I drive it twice a day taking the kid to schoolThis is what I don’t get. Am I misunderstanding something? It is my “assumption” (I know, I know) that there are supposed to be improvements albeit small iterations each time on all these areas. Or even SOME of these areas. Why is it that there are so many regressions in most if not ALL these areas with each update. It seems like there is either no progress or that we are moving backwards in most areas.
Ski
These point to map problems.Approaching a T junction stop sign where the main road is curved. FSD ignores approaching cars from left and tries to turn right in front of them.
Next T junction where we have right of way, no yield, just a right turn. FSD still comes to a full and complete stop and edges forward before make an extremely slow turn. I usually take this corner at 20mph.
Next is the left turn at the 4way where we have right of way (no road signs). FSD comes to a complete stop again before trying to turn left into someones garden, then overcorrects and cuts the entire oncoming lane to get to the correct lane.
Then another left turn T with full right of way, another full stop and stupidly hesitant turn with lots of stops and wheel spinning.
All that in the space of 400 yards. Every single one completely useless.
True - kind of “edge case” to have T-junction at a curvature.Pulling out in front of vehicles on curved roads leading into T intersections might not so much be a mapping problem, I think the system has trouble interpreting things on curved roads in general when compared to more perpendicular movements i.e. how the vehicle brakes when vehicles approach in the oncoming lane on curved 2-lane roads.
Trajectory in a curve is probably much more difficult to extrapolate with a high degree of confidence