One other thing to note, I really need to get use to my default disengagement being to move the gear shift lever…to many times times I take over the steering and get surprised that cruise control is still going
Last edited:
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Wiki MASTER THREAD: Actual FSD Beta downloads and experiences
- Thread starter jebinc
- Start date
EVNow
Well-Known Member
Yes, chances of an accident by 3k people in a week is > chances of an accident by 3k people in a week. But, we are looking at few months of testing by 2k people vs 1 week of testing by 3k people. Math is not daunting.
Now it seems - they will rollout to new people with every release - once in 2 weeks ? May be ... we'll see. I do expect them to stop at some point - may be after they reach 10k (or 20k) - whatever is their target. I also expect this fsd beta early access group to continue for a couple of years until it goes GA beta.
scottf200
Well-Known Member
It was Aug 1, 2017 (Autopilot, Processors and Hardware – MCU & HW Demystified – TeslaTap)
I mentioned mine was AP2.0 in June.
One of the hardware diffs was the radar but that is irrelevant AFAIK now. First I heard that cameras would make *any* difference? Source??
Aside: I did replace my AP2.0 repeater (fender) cameras to the latest [more] 'colored' version but not the tri front cameras.
I see Model S and Model X 75D with FSD OTA (2021.32.25) on TeslaFI. They were certainly AP2.0s. [update: maybe not looks they they were a mix of AP2.0 and AP2.5!)
Last edited:
One thing I am seeing FSD Beta consistently do in neighborhood driving or any driving without a clear lane divider, it will tend to drive down the center of the street. No bueno.
Lowchrg
Beta Blue-balled
It was Aug 1, 2017 (Autopilot, Processors and Hardware – MCU & HW Demystified – TeslaTap)
I mentioned mine was AP2.0 in June.
One of the hardware diffs was the radar but that is irrelevant AFAIK now. First I heard that cameras would make *any* difference? Source??
Aside: I did replace my AP2.0 repeater (fender) cameras to the latest [more] 'colored' version but not the tri front cameras.
I see Model S and Model X 75D with FSD OTA (2021.32.25) on TeslaFI. They were certainly AP2.0s. [update: maybe not looks they they were a mix of AP2.0 and AP2.5!)
How do you figure that they’re a mix of AP2.0 and AP2.5? It seems almost every 2017 that I have spoken to that received the beta is a late Aug 2017 or later Model S with AP2.5.
There is currently no evidence that suggests a single AP2.0 vehicle has gotten access but I would love to be proven wrong and give some hope to my vehicle that’s going on 3 weeks at 100 score with no beta to show for it.
AlanSubie4Life
Efficiency Obsessed Member
I didn't say math was daunting!
And I'm not sure I am following what you're saying.
I'm still not following. Perhaps you can clarify what you are arguing? (I think you're saying there were 2000 testers for several months in Tesla (we wouldn't know about their accidents BTW) and now we've added 1k more?)
If you use Tesla's PCF of 0.682854 accidents per million miles which yields a score of 100, and you assume each driver is driving about 250 miles a week (13k per year), and there are 2000 100-score public participants in the Beta.... (I'm using 2000 because it seems reasonable that it's considerably more than the 1000k Elon originally mentioned.)
0.682854 collisions/million miles * 250miles/wk = 0.000171 collisions/wk per driver. I think you can think of this as a 0.0171% chance per week? I don't think that's quite right since a driver could have more than one accident in a week...but whatever. Or a 99.98% chance of no accident in a week.
Anyway, let's just use that assumption, 0.017% chance of an accident in the given interval of a week:
With 2000 participants, that means (1-0.000171)^2000 = 71% chance of no collisions in a week in the 2000 public beta testers. That's not awesome. But it's also not horrible.
Could be wrong with this simple math (probability is hard - there, now I said math is hard...
). Someone can correct me if I've screwed it up. Seems kind of reasonable.Obviously each owner isn't going to be using FSD Beta all the time, so maybe lower the mileage to 50 per week. That brings the chances of the desired outcome up to 93% if we're only counting accidents with FSD Beta engaged. Obviously this assumes risk of accident on FSD Beta is the same as the risk when not using it (which is invalid but I don't know what is the relative risk).
Last edited:
As of now, I have seen the Model 3 steering wheel turn far faster than I have in a yoke video.
That's a great idea, and I'll be doing it from now on.
HA! In that Tweet, Elon refers to it as a YOKE!
IMO, that means you are doing it exactly right, keeping your eyeballs glued outside. I know, I know... I sound like a Dad with all of my soap box pronouncements of being safe with this thing, but you guys have gotten to be like extended family, and I don't want you getting hurt! <sniff> So keep your eyes on the road, damnit!
On to more pressing matters, like the "yoke" terminology:
By the authority vested in me by the Old Pilot Rules and the State of Confusion, I hereby pronounce and decree that use of the term "Yoke" when used to describe the vehicle steering apparatus is hereby approved and encouraged.
I have spoken.
There. Matter settled.
On a more serious note, I did have an interesting day with FSD. I drove through a part of town with quite standard roads and markings, and was able to go an impressively long distance without having one intervention.
One occurrence I did want to run by you guys. I'm stopped at a traffic light. No one in front of me. I'm in a left turn lane, with my left blinker on. There's another left turn lane next to me (on my left), and two straight lanes to my right. The traffic light is configured with two left turn arrow lights, and two straight round lights. I did not have FSD engaged. When the two straight lights turned green, my car gave me the "hey dummy, the light is green" ding. The two left arrow lights are still red.
What do you guys think? If I'd had FSD on, do you guys think it would've taken off, performing the left turn against the red light? I considered engaging FSD at that time, but chickened out.
I'm just wondering how the green light ding and FSD work things out between them.
I don't know if this is necessarily no good. It gives more reaction time if a pet or child runs into the street. The problem I have is that it doesn't move over soon enough when another car is coming toward you.
Last edited:
The lights that the car thinks are controlling should be bigger and brighter on the visualization than the other lights.
I only occasionally glance at the display, and honestly I couldn't tell you which lights were bigger/brighter on the display when this happened.
Darn it, I need my wife to ride shotgun with me on these rides so she can tell me what's going on with the visuals.
Well, this evening I was driving up a hill and there it goes in the middle of the road. The big problem here you can't see if there is incoming traffic because of the crest of the hill...and even when it's flat and there is an incoming car, it is a freaking road hog and will inch over at the last second leaving the other car a few crumbs...and that is when it is being nice, typically it will lurch to the right and hard break to a snails pace at the last moment.
yeah, i'm pushing that button fast and furious. would be great if there is rapid improvement on these things. seems like the correct behavior should be to tend to stay on the right on any road that doesn't have a divider, not default to the center.
On major stuff, I double down and send an email at the end of the day - per the instructions we received in that email stating our car would be seeing the beta download - just before it started downloading.
“How to provide feedback:
- Press the "Video Record" button on the top bar UI to send an Autopilot Snapshot video clip.
- Clips are automatically sent to the engineering team. You will not be able to view the clip.
- You can email your feedback to fsdbeta@ Tesla dot com
- In your email please include date, time, location, and if you took an Autopilot Snapshot. This helps us investigate issues, and better understand your feedback”
I haven't seen this behavior. It usually is late to move over which causes the other car to slow down. Once it moves over, it continues to drive at it's original speed. If there is a car parked along side it will sometimes slow down.
Last edited:
It used to do this, and people complained. When there were cars park along side the road, the car would serpentine down the road. No car, move to the right. Car, move to the left to go around. So they changed it to stay just to the right of the middle of the lane.
I would probably keep one hand on the yoke at all times inside one of the “loops” and basically rotate my hand/arm with it while it spins. Less than ideal though, certainly. I’d also probably be overriding via the steering mechanism because that would be nerve wracking.
Agreed. Most human drivers in my area do the same thing: drive in the middle of the street until they see an oncoming car.
I don't know if they have tamed the yoke, but there is no way to keep up with the speed of rotation of the steering wheel on my Model S.
Similar threads
