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.
I just got the update and went around a bit also. It looks the spot has to be pretty tight for auto park to show up. I was able to do it twice, but didn't try going down the same street again. Honestly, the spaces it was able to do are the only ones I'd particularly care about having it. If there's plenty of room, I can do it myself.Tried the new update - I dont see a good reason to use AP on surface streets at 35 mph unless you are stuck in traffic. It does pretty poorly on turns that has signs to slow down at >15 mph . Effectively an almost 90 bend. I dropped the Tac speed less than 20 mph the car kept driving straight. What's needed is 75 mph Auto Steer on freeways.
Auto Park is like winning the lottery ticket. I kept going around my neighberhood and was able to enable it ones. When it did show the "P" sign it did an amazing job parking the car. I wouldn't have parked the car in the spot thinking it was too small for the MS. It did a buitfual job parking, However when I tried doing it a second time around at the same spot - no luck. I recall AP1 being pretty bad during test drive too.
I had another idea driving home tonight, and maybe this had been suggested before.
When using autosteer, if I feel that the car is getting too close to the right (or left), I can apply opposite turning pressure to the steering wheel, but not enough to disengage AS, but rather, give autosteer a subtle "hint" that I want it to nudge ever so slightly in the direction I'm tugging on the wheel, but of course, still maintain lane tracking. When I stop the pressure, AS can return to it's normal activity.
Now, I have to turn the wheel to disengage AS, pass the car that's VERY close to me, and then re-engage. It's a real hassle.. I'd love to just give AS hints as to what to do in tight situations without disengaging it.
AP1 had its issues. Do some searching in the 7.0 threads in the October/November/December 2015 timeframe for details when it first became available.anyone know if AP1 was this bad when it was initially launched? Im looking for a gleam of hope that the system will get better and that its not some inherent flaw with the revised hardware and/or Tesla Vision. I dont mind the wait (to a certain degree)
local roads autosteer is straight up sketchy when i tried it this morning. Unusable for me at least.
AP1 was fully functional when launched
AP1 was fully functional when launched
Is this sarcastic? It was definitely not. The first time I test drove 7.0 with ap1 we ended up nearly sideways on a freeway exit and the Tesla employee in the car screamed like a girl!AP1 was fully functional when launched
Playing with it a bit more, it's pretty easy to tell where it'll consistently fail. The big one is on turns. Regardless of the direction of the turn, it tends to drift too far towards the outside, with the degree to which it's off increasing the sharper the turn is. This makes it almost pointless, since TACC is more than capable for straight roads.
The main other failure is a bit more reasonable: it's not great at negotiating when lanes split or merge. On a lane split, it seems to go deer-in-headlights until drifting far enough into one of the two lanes and sticking with it(again, reasonable). On a lane merge it tends to swerve to try to stay in the middle *and* the right.
Definitely not ready for prime time, but I'm sure it'll improve. Auto-park works like a dream, though, assuming you can find an applicable spot.
I'm starting to feel like the shadow mode is just a shady cover for the delaysEach time I engaged autosteer on surface streets I had to take over within seconds because of its last to drive into the adjacent lane despite blue bars showing around the current lane.
My wife was asking me why release it in this way rather learn from shadow mode and then release it, I didn't have an answer. The neural network should be able to train itself using shadow mode data too, is there a reason why it has to be live to get better overtime?
The NN is trained, once, using the same type of data that it needs to process. Its purpose is either to identify features in live video, or to command driving based on live video, so it is trained using a curated set of video frames. This happens at Tesla HQ and the result is uploaded to the cars - it does not retrain itself based on driving data.The neural network should be able to train itself using shadow mode data too
The first try looked OCD.Here is an AP2 AutoPark Video