Auto-park fail - manually stopped just in time

[vadimr]

The auto-park feature is worthless junk that probably was written as part of some hackathon and forgotten. The same thing on 2016 VW GTI works 10x better. This POS hardly ever detects any spaces and when it does they are all wrong. I often back-out from my garage and this idiotic thing thinks I want to perpendicular park between a tree and basketball hoop. Couple times it offered to park me on a lawn when I was just parallel parking it was like hey look at this fancy spot on the lawn in the middle of the street. I thought it was going to parallel, nope the useless crap was backing up into the curb and straight onto the lawn.

This is why FSD is vaporware, it will never come out. Those of you who paid 3k just threw your money out. Full self driving? It hits a stopped firetruck! Yea yea sure, it will fully drive by itself. Waymo is like 10 light years ahead of Tesla in addition to having LIDAR and even they are not ready for a release but Tesla with their fancy cameras will have this in a few months.

I love my Model 3 and I love the autostreet and it's pretty darn good but no way we are ever getting FSD on these cars. Even the current EAP when it goes around the bends with a car in adjacent lanes which it sees just steers right into them. I can reproduce it literally every single time. Yes humans have the tendency to not be perfectly in the middle of their lane when going around bends but the logic should account for that. For example: I have a car on left and we are in the right bend, that car is coming way too damn close to me, let me move to my outer most limit within my lane or slow down to let it be out of danger. Nope, this thing just continues on the collision course.

 

[Moderatefan]

You missed his tweet from over a year ago

I did indeed, but don't see what good it does to anybody on this forum to bring up obsolete estimates when new ones were provided.
We do know things change, so its best to move on.

 

[LostInSpace]

Had the same thing happen to mine. Autopark is dead to me.

Mine hit a curb but with the rubber fortunately ... not the rim. I guess it can't sense too low or too high but just right ...

 

[pipplo]

My problem with the 'driver must be aware' is that you never really know when to intervene for sure. If the op stopped it at 20cm who knows if the car was going to stop at 10cm?

It's impossible for the driver to know what the car is planning to do so then you are stuck in this guessing game of 'should I stop it or not'.

 

[flashflooder]

My problem with the 'driver must be aware' is that you never really know when to intervene for sure. If the op stopped it at 20cm who knows if the car was going to stop at 10cm?

It's impossible for the driver to know what the car is planning to do so then you are stuck in this guessing game of 'should I stop it or not'.

Precisely the reason I never use it. Using it is more nerve-racking than just parking it myself.

 

[Barklikeadog]

It does say in the manual that you are responsible for it.

The driver has been warned in the manual, and the placement of the ultrasonics is fairly limited, there's not a whole lot that can be done. It's also an assistance feature meant for human supervision, it's not a fully autonomous feature.

You're not one of those guys who takes their hands off the wheel during Autopilot, right?

FSD should be turned on soon where the car drops you off and seeks out a parking spot.
lol

 

[SigNC]

Auto park seems like a gimmick IMO. I'd much rather park myself and be done with it.

 

[chronopc]

Auto park seems like a gimmick IMO. I'd much rather park myself and be done with it.

I wouldn't say it's a gimmick. It's an option. A lot of people prefer to just park themselves and they should. With how tight the spaces are where I live, I have found the feature really useful.

 

[Moderatefan]

A lot of pessimism in this thread. I have no doubt we will get there eventually. The only issue is whether you want to be a beta tester and help the progress or you want to wait until 100% of kinks are worked out.
And I mean FSD in general, not the current ultrasonic thing.

 

[chronopc]

A lot of pessimism in this thread. I have no doubt we will get there eventually. The only issue is whether you want to be a beta tester and help the progress or you want to wait until 100% of kinks are worked out.
And I mean FSD in general, not the current ultrasonic thing.

I accept that I'm a beta tester whenever I purchase any technology. I'm still waiting for the kinks to be figured out on my phone. Lol

 

[vadimr]

A lot of pessimism in this thread. I have no doubt we will get there eventually. The only issue is whether you want to be a beta tester and help the progress or you want to wait until 100% of kinks are worked out.

You won't get there in this Model 3. I am willing to bet that FSD never comes to M3. Maybe some more advance highway driving but 0 chance to Level 4 or 5 autonomy. ZERO. Google had highway driving down to 2-3x safer than a human being about 10-12 years ago. They even stopped doing highway testing because they succeeded to be significantly safer than a human being. However they always said highway was super easy. Tesla didn't even finish that phase and they are talking about FSD? Something that Google still cannot finish with much better resources and LIDAR? Come on, I like Elon and I like his way of pushing the envelope and trying to break barriers but this *sugar* is total vaporware.

The mistakes I see my EAP make daily are super dumb, that either means it is not smart enough to process those operations or too slow. Regardless of what it is, it's not working. I mean come on, it gets scared of a shadow. Driving under an overpass it freaks out. Literally scared of it's own shadow On country type roads coming to a stop red light behind another vehicle where suddenly there are 2-3 lanes down from 1 it was driving it flips out and starts going into random directions. I had it go all the way right and suddenly flip out and go all the way left. FSD? Oh yea sure... They have a nice fancy video of that FSD in action on a pre-recorded road tested 300x.

 

[JeffK]

You're not one of those guys who takes their hands off the wheel during Autopilot, right?

FSD should be turned on soon where the car drops you off and seeks out a parking spot.
lol

Not sure what you're referring to, perhaps you hit reply on the wrong post.

Personally I don't think FSD comes into the conversation with the current implementation the auto park assist since it clearly is not mapping the environment ahead of time. One of the most important things with FSD will be how it maps the environment it's in plus it's own knowledge of "self" and where its boundaries are.

I don't think it's impossible with the current hardware. Except that with the current camera placement, if you map a location to park, the situation cannot change after the vehicle commits since it's out of the camera view (unless ultrasonics catch a new obstacle). Downward facing cameras would be able to use vision to update maps in real time combined with data from the ultrasonics instead of relying on a map that's now stale after several seconds.

That said, humans make a mental map before they commit and do this all the time without a great view of where the curb is or if a small child were to run into the spot etc. I think with the current camera setup they can still do a pretty good job if they decide to have it map the environment first.

 

[Moderatefan]

You won't get there in this Model 3. I am willing to bet that FSD never comes to M3

I can't guarantee that current M3 hardware is adequate for FSD, but it doesn't mean it's never going to happen.

Current EAP is definitely not representative of what can be achieved in the future.

I'm taking a Machine Learning class, cause I want to understand what it's about. I'm like 30% into it and just touched on neural networks, so I don't yet have a full picture. But it does seem very promising. It's basically re-creating brain function by training a "programming representation" of it on a number of scenarios, the same way it happens in life. So, whether it's walking, or driving, or listening to a music, or predicting what's going to happen - you basically accumulate a number of experiences in your life, classify them into these specific categories and then based on the data you accumulated in each category, you learn to recognize/classify/predict.

Driving is just one of many tasks that brain deals with, so you can limit the data you need to feed to the model to this subset of experiences.

It is still a gargantuan amount of data that needs recognized/processed by the model. But you can see how your model is doing in recognizing the data you feed to it and make adjustments, so it becomes better at recognizing your training data.

I think this approach has a good potential in replacing the human at these tasks. But will see what happens.

 

[chronopc]

You won't get there in this Model 3. I am willing to bet that FSD never comes to M3. Maybe some more advance highway driving but 0 chance to Level 4 or 5 autonomy. ZERO. Google had highway driving down to 2-3x safer than a human being about 10-12 years ago. They even stopped doing highway testing because they succeeded to be significantly safer than a human being. However they always said highway was super easy. Tesla didn't even finish that phase and they are talking about FSD? Something that Google still cannot finish with much better resources and LIDAR? Come on, I like Elon and I like his way of pushing the envelope and trying to break barriers but this *sugar* is total vaporware.

The mistakes I see my EAP make daily are super dumb, that either means it is not smart enough to process those operations or too slow. Regardless of what it is, it's not working. I mean come on, it gets scared of a shadow. Driving under an overpass it freaks out. Literally scared of it's own shadow On country type roads coming to a stop red light behind another vehicle where suddenly there are 2-3 lanes down from 1 it was driving it flips out and starts going into random directions. I had it go all the way right and suddenly flip out and go all the way left. FSD? Oh yea sure... They have a nice fancy video of that FSD in action on a pre-recorded road tested 300x.

FSD is decades away for sure.

I mean most humans can't even figure out how to drive.

 

[JeffK]

FSD is decades away for sure.

I mean most humans can't even figure out how to drive.

haha I don't know about that, I think human level FSD is right around the corner. 100x better than a human might be decades away, but not doing bonehead things should be pretty easy.

 

[Thomas Edison]

Has this happened to anyone else? Kind of disappointed that my auto park is only at a 50% success rate.

50% success rate with a sample size of 2 attempts . While technically true... come on man!

 

[Moderatefan]

haha I don't know about that, I think human level FSD is right around the corner. 100x better than a human might be decades away, but not doing bonehead things should be pretty easy.

It is also easy to see why Elon is so afraid of AI abuses. FSD is a manifestation of AI in a limited area. If you combine a bunch of these little AIs into one or create a super smart model that can accept any sort of training data and learn from it, then you've exceeded human capabilities, b/c a human cannot compete with all the processing power available in the cloud infrastructure even today.
Then the question becomes whether some monopoly like Google abuses it to the point of decimating all competition or whether AI can become "self-aware" or, alternatively, decide to act on its data in the way it sees fit w/ disregard for directions from humans w/o having far reaching conscious goals.

 

[JeffK]

It is also easy to see why Elon is so afraid of AI abuses. FSD is a manifestation of AI in a limited area. If you combine a bunch of these little AIs into one or create a super smart model that can accept any sort of training data and learn from it, then you've exceeded human capabilities.
Then the question becomes whether some monopoly like Google abuses it to the point of decimating all competition or whether AI can become "self-aware" or, alternatively, decide to act on its data in the way it sees fit w/ disregard for directions from humans w/o having far reaching conscious goals.

A little bit of a tangent from autoparking (should move it to its own thread). But, you don't have to exceed human capabilities at all to make evil AI. We could have evil AI right now that discriminates against certain people w.r.t loans, accepting people into college, sorting job applications, or ones that generates fake news. What's needed is regulation in this space to make sure it's acting fair like a human is supposed to according to the law. Make sure AI, or any machine learning algorithm for that matter, is not being evil (even if it's statistically correct ).

 

[Nuclear Fusion]

I did indeed, but don't see what good it does to anybody on this forum to bring up obsolete estimates when new ones were provided.
We do know things change, so its best to move on.

Even better to stick your fingers in your ears & sing la la la

 

[insaneoctane]

I wouldn't say it's a gimmick. It's an option. A lot of people prefer to just park themselves and they should. With how tight the spaces are where I live, I have found the feature really useful.

I'd call an option that doesn't work a gimmick.