Tam
Well-Known Member
Navigation on Autopilot would stop your car at the end if it is not already following a lead car and if there's no intervention from driver:
It would prompt:
"Autopilot navigation complete
Press Accelerator to resume"
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Accident while on EAP...
- Thread starter Freewilly
- Start date
Navigation on Autopilot would stop your car at the end if it is not already following a lead car and if there's no intervention from driver:
It would prompt:
"Autopilot navigation complete
Press Accelerator to resume"
How did your friend get Tesla to look at his data and escalate it internally? I've contacted Customer support and all I got back was sorry, glad everyone is OK, reach out to your local service center for repairs... I want them to look at the data and hopefully find out if there is a software issue or not.
He presented the issue as a failure of AutoPilot, and the cause of the accident. He said Tesla took the matter seriously and escalated it immediately to the AutoPilot team. He did email a description in detail, including video, and Google Map Data. He followed up with a phone call, and then the Tesla team got engaged. On the other hand, they still did not get back to him with the conclusion, nor did they offer to help cover any costs. You might want to try again.
There is some interesting research that is being applied in this area "monocular depth estimation". I've been watching some of it this evening. Its important to realize its still an estimation. Have a look: Papers With Code : Monocular Depth Estimation
Enginerd
Member
This is true, however we know that in V9 (?) software, one of the ways they're processing the image data is to run the neural network algorithm on 2 successive frames instead of just one. This way you get a rate of change of size/distance from processing a single pair of images. I think this was one of the conclusions drawn by @verygreen . So, the system is capable of extracting more than you may be initially giving it credit for.
Daniel in SD
(supervised)
The human brain is an amazing thing
I think Elon's argument that FSD can be done with only cameras because humans can do it with a single eye has put Tesla on a very challenging path.
It might stop at the end if you reach the end of the route.
But it most often drops out of NoA and transitions to AP. It will pop a very small message. But I would not call it a nag (does not vibrate wheel or flash screen or make a noise ahead of time). You can easily miss it.
But it will give a gentle jingle of some sort after it drops to AP.
It did it on me tonight. 99% of the time it transitions to AP because NoA rarely ever stops/ends at your destination.
There are cases where AP ends and it says take the wheel etc. It's usually AP that "Ends". NoA usually transitions to AP long before it.
Sometime it does go from NoA to nothing.
sperkin
Active Member
I experienced this before too, but I slammed the brakes myself. My experience is always with white cars though. I have also experienced once a small blue hatchback that was invisible to the side radar and camera. The car almost auto lane changed into it.
Most of the time it’s the front camera and radar not seeing white cars. Just today a White Model 3 was completely invisible to the radar and camera. Almost rear ended him when he changed lanes in front of me.
Most of the time it’s the front camera and radar not seeing white cars. Just today a White Model 3 was completely invisible to the radar and camera. Almost rear ended him when he changed lanes in front of me.
AlanSubie4Life
Efficiency Obsessed Member
TeslaVision is probably so good now they have started ignoring the radar. They don’t even need it any more, it’s like a child’s toy! This is because of the huge leaps forward in the neural net training.
Unfortunately there are some minor edge cases.
sperkin
Active Member
Maybe it's so reflective that the radar doesn't see it. I don't know. Always nice shiny white cars in the morning with the sun behind me is where I see this problem. It's hard to show on camera because the car is doing 1 to 4 mph but it doesn't try to slow down or stop for white cars. Sometimes it'll even try to speed up. Very subtle but I have hit the brake to cancel AP. Don't really want to test it out.
where did you get the graphics from?
And which organization conducted the test? the IIHS?
I can’t tell if you’re being sarcastic; but I can tell that Tesla still relies on radar input and will immediately abort AutoPilot if it experienced a reduced radar condition. It will also not allow AP to re-engage without it.
AlanSubie4Life
Efficiency Obsessed Member
I was being sarcastic.
This whole conversation makes me a little sad. In this case it was probably driver override of AP that caused the problem, but AP has other known problems, and I feel like this FSD thing is really far off at this point, and while it’s fine (essential?) to be working on it, Tesla should be more focused on promoting their cars because they are awesome just as they are. They’re so good you would not want a computer to have all the fun driving them! That seems like a better strategy to me. It’s not like they are making tons of money off of FSD - so why hype it so much when it does not actually exist in a workable form, and likely won’t for quite some time?
There isn’t going to be anyone reading a book nonchalantly in a Model 3 while it drives itself for at least 4-5 years is my feeling. Well, there may be, but I mean accomplishing this without reduced life expectancy is unlikely in that timeframe.
Last edited:
I would not have bought the car if not for FSD. And I know my friend bought it for the same reason. And I didn’t expect it to work for 4-5 years and I know it will have limitations when it does.
Similar threads
