-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Nope, it's not ready
- Thread starter Beta V
- Start date
-
- Tags
- auto drive
powertoold
Active Member
We know it uses open sourced maps for exits and interchanges. When people have a problem with NoA and exits, they look on the open source map and see that it's incorrect.
I don't think you understand my point about technology local maximum. Did you watch the autonomy day and also Elon's comments about "local maximum?"
Knightshade
Well-Known Member
Nor is it supposed to be, or intended to be used there.
Man, if only this was mentioning in the manual or something, right?
Oh, wait... it is.
So, user error then. Since, again, it's not supposed to or intended to work there
Tesla: City driving is coming at the end of the year.
You: What's wrong with this car, why doesnt it drive in the city like I think it's supposed to?
Tesla: (facepalms)
powertoold
Active Member
I didn't come up with it. Elon talks about it often when he refers to lidar and HD maps.
My point is with the most CURRENT Tesla software and hardware the car is not safe for in city AP. Hopefully city driving software update by end of this year will be safe. Read my post Once again. You haven’t once tried AP in city streets and only used AP on highways? If true, you must be very honest and rare individual. Be honest with yourself and stop the sarcasm.
I agree with you, VM.One. For the most part it is good both highway and city but not safe enough in city YET with the current software. Folks on here are tell others to read the manual. I’m only sharing a few experiences of city driving with the most current software and hardware from tesla and in my opinion it’s not ready yet.
Knightshade
Well-Known Member
And mine is that it's not supposed to be.
And the manual explicitly tells you that.
So your "point" really... isn't one.
Why announce your car does not do something it is not supposed to do anyway?
What's the point of that?
It depends what you mean by "safe"
Anyway Elon has said there's going to be 3 stages for FSD-
1) feature-complete: this, they are hoping, will be out to at least a small # of early-access folks by end of this year. NOT wide release.
This means the car can do city driving, under human supervision, and likely still requiring intervention some of the time, but not most of the time.
It's unclear when this will go into general release.
It's also unclear if you, personally, would find this "safe"
2) feature complete and safe without supervision: this is when it works to a degree Tesla no longer thinks it needs to be supervised....but still will officially need to be for legal reasons.
3) feature complete and regulators agree with Tesla that it's safe enough to not require supervision. Again no known date for this.
The difference would be I understand it's not supposed to work or be safe there and thus don't waste time complaining about something not working that's not intended to work.
They don't even have speed sign reading yet. Reading arbitrary traffic signs is likely a long way down the road, so to speak.
Which is fine, and its good ppl are trying it in different environments. The push back is for people who drive it in the city, expect it to work, and then complain loudly about how awful AP is “not ready”, “doesn’t work” etc etc. The fact is, AP is a beta work in progress, and should be treated as such. When asked, I equate AP to a responsible teenager who has just passed their driving test: cautious, anxious, but not sure how to react in unusual situations.
I think it works well on city streets. I fully expect to take over at times. The way I look at it is per the manual I am fully responsible for the control of the car so it doesn't matter if I use AP or not on city streets. I am still the one responsible.
However looking at how it performs it will have to be a huge update for them to get to where they imply they will be before the end of the year. I would expect current software to be a lot closer than it is. I wouldn't expect a huge improvement all at once. The ability of it to follow lane lines I expect gradual improvements over time not huge improvements in a single update. If they are as close as they are implying why haven't they gradually improved it over time.
However looking at how it performs it will have to be a huge update for them to get to where they imply they will be before the end of the year. I would expect current software to be a lot closer than it is. I wouldn't expect a huge improvement all at once. The ability of it to follow lane lines I expect gradual improvements over time not huge improvements in a single update. If they are as close as they are implying why haven't they gradually improved it over time.
Knightshade
Well-Known Member
It matters in the sense that any time it "works" on city streets is pure accident.
The inherent assumptions of the software are:
All traffic is going the same direction
All access on/off the road is limited via ramps
There are no intersections and especially no cross-traffic (ignoring this is what killed both idiots who drove under tractor trailers on autopilot)
Using it when those things aren't true means you're just randomly hoping none of those things come up on roads where they inherently come up.
It would have to be an update that actually is intended to handle intersections and cross/oncoming traffic.
Which the current SW isn't.
What the current SW does in those situations tells you literally nothing about the ability of future software in those situations.
Which is a mistaken expectation, since nothing in current code (other than the 'warning you are about to run a red light') is relevant to city driving.
It's not that "their current city driving code isn't great"
It's that "their current city driving code does not exist"
It will in a future version- and again what todays SW does tells us nothing at all about how well it'll work.
How do you gradually improve something that doesn't exist?
City driving code is for HW3 using a much larger, more advanced, neural net.
Everyone, even HW3 cars, are still running the HW2.5 "only understands limited access one-way highways" AP code.
I had the free 30 days trial of FSD at purchase and at that time running on V9.0 software. I stopped using NOA because it was honestly unsafe and more stressful to drive on freeway. After the trial period I decided not to purchase and wait for the software improvement to justify FSD. V10.0 was released and I pulled the trigger a month ago. At that time NOA was not that great either so I didn’t use it at all. Not until 36.1 was released that made a huge improvement on NOA. I used today driving from San Diego to Orange County. Boy I was impressed! It was 90% better and for once in a long time I felt safe using it for most of the trip. Point being, the software improved over time and folks can see the difference and have confidence in the product . That’s how it should be so we have positive feedback from the software update. I totally agree that it shouldn’t be an on/off switch to be able to feel safe using AP on city street. Improvements should be incremental with latest software updates. Like I said in my original post, hopefully the future update will give us a sense of security while using it. I’m starting to feel that way about NOA on freeway now. Kudos to Tesla for the recent update.
MichaelP90DL
Active Member
You're quite a good writer. As for Angel...I had a girlfriend much like that. I can provide the number of her psychiatrist.
Kiwiadventure
Member
Don't use Auto Pilot as you understand it's in Beta testing and you signed the acceptance understanding you will accept is limitations. No vehicle applies the rear brake when using gears to slow down.
No you are wrong. I am not hoping none of those things come up. I am fully expecting those things to come up and am ready to take over. It isn't any different than if I was driving manually I am ready and able to react to conditions as needed. I just am not doing the fine lane maintaining and speed corrections manually. Most of the roads I drive on are divided though not limited access.
It reacts to cross traffic that is one of the things I override frequently. It tends to brake too much as most times the car will be clear well before I get there. However since it doesn't support cross traffic at this time it isn't even something I think twice about. I understand why it is braking for the cross traffic and consider it normal and expected for current software. I am happy to see that it is identifying the cross traffic as that means now they just have to teach it that it may not need to slow down. Right now it is assuming it is a stopped car in my lane and reacts correctly for that.
My bigger issue is how often it can't maintain the lane. I'm not talking about going through intersections either as I would understand that not working. Stretch of road divided 45mph limit no cross traffic and it gets confused on the lane markings. Another section with just TACC on it freaks out and has activated lane departure corrections even though I am in my lane and not close to crossing the line. I wouldn't even attempt that area with AP activated I have no expectations of it working. If they are able to run 3.0 specific software in the background for data gathering and testing then maybe they are as closer than I am giving them credit for. At this point that is the only way I can see them releasing outside of employees this year. If they can't run the 3.0 software in the background it will take months to reach parity once they release 3.0 specific code.
Not once have I seen the Red light warning. Not sure if it is because I don't run red lights or if it doesn't work.
NoAP misses the exit frequently. It keeps going straight though the exit curves to the right. I don't even use NoAP anymore, I will test when a new software release come out to see if it has improved but so far not enough for me to use it. This could be because how the lanes are marked around here I know other states have better markings.
At one point it used to detect traffic cones, it doesn't anymore and hasn't for some time now.
How far back it will stop is another issue I see. It really needs 2 settings 1 for follow distance and another for stopping distance. I don't need that much room between me and the car in front if we are stopped. This is easy to correct though and I could see it coming out in a software release. It also needs to be easier to use the throttle to ease up on the car in front if I want to get closer. Often done to allow the car behind me to be able to exit or access a turn lane.
Knightshade
Well-Known Member
Yeah either you have horrendously marked roads or there's something wrong with your car.
I have 0 of those problems- ever- and I'm on AP/NoAP the overwhelming majority of my miles.
Once in a while it'll tell me a given exit is unsupported-and do so far enough in advance for me to handle exiting- but it's never just flat out missed one- nor does it ever have trouble staying in lane when the lanes are actually marked.
Similar threads
