Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Will L5 FSD be possible with Model 3's AP 2.5 hardware?

Will L5 FSD be possible with Model 3's AP 2.5 hardware?


  • Total voters
    102
This site may earn commission on affiliate links.
Personally I don't think that geofencing small areas like college campuses or shopping malls for FSD meshes with Musk's numerous statements on self driving.

I also don't think that the hardware on the Tesla is capable of L4 within a geo-fenced area in all weather conditions.

L4 with geofencing is a crutch for regular drivers and, in my opinion, not particularly useful at all which is why I believe that most manufacturers are planning on skipping over it entirely.
The geofenced area doesn't have to be that small. Ford gave some examples in the article I linked:
“You could imagine a geofence could be Manhattan, it could be southeast Michigan with the corridors to the airport plus the major Metropolitan areas,”

The geofence can serve as a control on weather conditions (for example it never snows in certain areas).

And I should note that although Elon talks a lot in various places (not all of which are directly related to the current hardware, in some talks he refers to the direction Tesla is eventually going), the order page is really what we have to go by.
 
Personally I don't think that geofencing small areas like college campuses or shopping malls for FSD meshes with Musk's numerous statements on self driving.

I also don't think that the hardware on the Tesla is capable of L4 within a geo-fenced area in all weather conditions.

L4 with geofencing is a crutch for regular drivers and, in my opinion, not particularly useful at all which is why I believe that most manufacturers are planning on skipping over it entirely.



Technically, even Elon's dream of a coast to coast trip wouldn't even qualify as L5, as it would need to be geofenced to remain within range of the "robot snake" chargers.
 
Didn't know AEB on AP 2.5 works so differently with a Tesla. AEB in our Toyota just brakes if you get too close though I've only noticed it get activated twice.

Tesla is adding automatic emergency braking to ‘Autopilot 2.5’ cars with new update
“Automatic Emergency Braking, a new collision Avoidance Assist feature, is designed to automatically engage the brakes to reduce the impact of an unavoidable frontal collision with another vehicle. The brakes disengage when you press hard on the accelerator pedal, release the brake pedal, or sharply turn the steering wheel.”
 
The difference here is between "thing's Elon has said" and what is actually printed in the Tesla description of the "Full Self Driving" feature.
It's on the website, not the exact words Elon used, but enough that a person can assume that their Tesla with FSD will able to be used for car sharing and ride hailing.

upload_2017-10-11_16-27-3.png
 
  • Informative
Reactions: voip-ninja
"EM: Yeah, so this is using only cameras and GPS. So there's no LIDAR or radar being used here. This is just using passive optical, which is essentially what a person uses. The whole road system is meant to be navigated with passive optical, or cameras, and so once you solve cameras or vision, then autonomy is solved. If you don't solve vision, it's not solved. So that's why our focus is so heavily on having a vision neural net that's very effective for road conditions.
CA: Right. Many other people are going the LIDAR route. You want cameras plus radar is most of it.
EM: You can absolutely be superhuman with just cameras. Like, you can probably do it ten times better than humans would, just cameras.
CA: So the new cars being sold right now have eight cameras in them. They can't yet do what that showed. When will they be able to?
EM: I think we're still on track for being able to go cross-country from LA to New York by the end of the year, fully autonomous."

Transcript above is from a TED talk in April this year. FSD is Level 5 without any doubt. But any twist from the car maker for any future FSD demo won't surprise me, like geofencing, slow motion, or parking lots at Superchargers. They can even put the entire FSD design team and the mothership in the trailer of their new semi truck. Just need to bring out the truck first.
 
"EM: Yeah, so this is using only cameras and GPS. So there's no LIDAR or radar being used here. This is just using passive optical, which is essentially what a person uses. The whole road system is meant to be navigated with passive optical, or cameras, and so once you solve cameras or vision, then autonomy is solved. If you don't solve vision, it's not solved. So that's why our focus is so heavily on having a vision neural net that's very effective for road conditions.
CA: Right. Many other people are going the LIDAR route. You want cameras plus radar is most of it.
EM: You can absolutely be superhuman with just cameras. Like, you can probably do it ten times better than humans would, just cameras.
CA: So the new cars being sold right now have eight cameras in them. They can't yet do what that showed. When will they be able to?
EM: I think we're still on track for being able to go cross-country from LA to New York by the end of the year, fully autonomous."

Transcript above is from a TED talk in April this year. FSD is Level 5 without any doubt. But any twist from the car maker for any future FSD demo won't surprise me, like geofencing, slow motion, or parking lots at Superchargers. They can even put the entire FSD design team and the mothership in the trailer of their new semi truck. Just need to bring out the truck first.

Five minutes of driving on magnesium chloride coated winter roads will leave the entire car other than the windshield coated with a mag-chloride and dirt film.

The Tesla cameras don't have washers.

I don't see how it will be possible with just the cameras.
 
  • Like
Reactions: Kant.Ing
I think Elon was coming from the perspective that if AP2 HW can provide all the input (visual data) that is available to human eyes, and has enough computing power ti process the data in real time, then it should be able to reproduce a human response. The main unknown in this assumption is what is the computation power required to process the data in real time, and how much of that is SW limited vs HW limited. I don't think we know at this point.
 
Agreed, I'm unclear how they can possibly keep anything other than the front sensors clear of snow and road salt. (Heating to melt the snow without wiping and adding washing fluid will just bake the salt on thicker for the cameras). There's going to be a lot of stopping at "gas" stations to wet and squeegee the sides and back of the car - that'll get old real quick during ski season.

Even ultrasonics don't stand up to the right kind of snow - I've had to listen to the continuous beep of impending doom for hours when stuck in near (but not total) gridlock on the highway in a big storm (traffic moving just enough to be totally unsafe to "park" and clear the sensors - and absolutely no way to pull off).
 
Agreed, I'm unclear how they can possibly keep anything other than the front sensors clear of snow and road salt. (Heating to melt the snow without wiping and adding washing fluid will just bake the salt on thicker for the cameras). There's going to be a lot of stopping at "gas" stations to wet and squeegee the sides and back of the car - that'll get old real quick during ski season.

Even ultrasonics don't stand up to the right kind of snow - I've had to listen to the continuous beep of impending doom for hours when stuck in near (but not total) gridlock on the highway in a big storm (traffic moving just enough to be totally unsafe to "park" and clear the sensors - and absolutely no way to pull off).

Almost seems like a situation where one shouldn’t be on the road at all, eh?

I always wonder about that- robots and AI still can’t beat physics AFAIK, so will we have to get used to the fact that, “This FSD thing sucks! I can’t go out in this freezing rain because my car can’t see or grip the road- WTF?! If *I* were the one driving, I’d be totally fine!” When in reality, “you” wouldn’t be.

Perhaps FSD will save us from ourselves?
 
  • Disagree
Reactions: voip-ninja
The first thing that needs to be fixed before FSD is the navigation. The navigation in a Tesla is not perfect, doesn’t understand detours, traffic is not always real time,and generally the turn by turn is not 100% reliable. If you don’t have almost perfect navigation you can forget about a self driving car. All the chips, sensors in the world aren’t going to fix this particular issue.
 
  • Like
Reactions: internalaudit
I´m seriously shocked that there are still almost 20% who believe Elons FSD blabbering despite Tesla`s track record of under-delivering, despite his own development leads not believing him, despite everyone else in the whole industry having completely different time plans and sensor setups.

I completely believe that Tesla will reach level 4-5 eventually with a Model 3 ....with the 2nd or 3rd iteration of the model/ap/sensor hardware....

We should simply ignore Elons extreme optimism and just look at the competition for realistic estimates...
Those estimate level 3/4 for highways and some urban areas in 2020-2021 and full lvl 5 autonomy at some later date (some say late 2020s)
 
Last edited:
The first thing that needs to be fixed before FSD is the navigation. The navigation in a Tesla is not perfect, doesn’t understand detours, traffic is not always real time,and generally the turn by turn is not 100% reliable. If you don’t have almost perfect navigation you can forget about a self driving car. All the chips, sensors in the world aren’t going to fix this particular issue.

I think that's mainly because the navigation is not yet relying on visual data. If the cameras can see the road sign for detour the car will just follow that. Right now navigation is mostly GPS data with delayed traffic information.

I suck myself at navigation and I'm quite far from 100% reliable when I get in unfamiliar area, doesn't mean I won't succeed to get to my destination. Just that it might take a little longer in the process. As soon as vision is solved the navigation is too.
 
Look up L5. It requires full redundancy. That is part of the L5 definition.
Actually it doesn't clearly state it, nor is it part of any L5 definition that I've seen. That misinformation is constantly quoted here as fact, and seems to be factually wrong.

I used to do it too, until I actually did the research and realized it's not easily found (if it at all exists), and for something that important they would make it more prominent.

I'd gladly be proven wrong, so please give me a link from a credible source (not a forum) that states that L5 requires full redundancy.

Some of my research:

There's only 1 mention of redundancies here, nothing interesting.
https://www.nhtsa.gov/sites/nhtsa.dot.gov/files/documents/13069a-ads2.0_090617_v9a_tag.pdf

There's only 2 mentions here, and neither give details (this used to be on the NHTSA website, I can't find the link now, might be the previous version of the next link)
https://www.fenderbender.com/ext/resources/pdfs/f/e/d/Federal_Automated_Vehicles_Policy(1).pdf

The preliminary concerns are they need to investigate what levels of redundancy are required, this hasn't been made a REQUIREMENT [yet?]
https://www.nhtsa.gov/staticfiles/rulemaking/pdf/Automated_Vehicles_Policy.pdf

Nor is there much about required redundancy here, which is the NEW US DOT release from last month -- September 2017 (latest stuff)
Automated Vehicles for Safety

The NHTSA adopted the SAE standards a while back, nothing about redundancy here
http://www.sae.org/misc/pdfs/automated_driving.pdf

There is a 30 page document from the SAE, which I don't have access to and don't want to pay for, which MAY mention required redundancy, so if you can post that, and show where L5 requires full redundancy, I'll agree with you. Otherwise please stop spreading misinformation.
 
Actually, Tesla's description of the "Full Self Driving" feature is NOT Level 5. Nowhere do they ever say the car will drive without a driver. They only say it will enable "full self-driving in almost all circumstances", and "be able to conduct short and long distance trips with no action required by the person in the driver’s seat".

Two words: "Tesla Network".
 
  • Like
Reactions: FlyingKiwi
Almost seems like a situation where one shouldn’t be on the road at all, eh?

I always wonder about that- robots and AI still can’t beat physics AFAIK, so will we have to get used to the fact that, “This FSD thing sucks! I can’t go out in this freezing rain because my car can’t see or grip the road- WTF?! If *I* were the one driving, I’d be totally fine!” When in reality, “you” wouldn’t be.

Perhaps FSD will save us from ourselves?

There are plenty of driving scenarios where the sensors are going to struggle but a human driver would not.

Scenario where you are getting road salt and grime baked on your entire car is a weekly occurrence around here for 1/3 of the year and I don't think most managers would accept the "well, I shouldn't be on the road anyway!" excuse when not showing up for important work meetings at the office.

A lot of industry people have been highly skeptical of Tesla's claims for exactly this reason. Some competitor platforms are going to rely more heavily on lidar coupled with the front sensors which can be cleaned when the windshield gets cleaned. Some are coming up with ways of washing the AP cameras with heated jets, etc.. I'm not aware of any that are doing what Tesla is currently doing and promising full autonomy.
 
Almost seems like a situation where one shouldn’t be on the road at all, eh?

And what pray tell is one supposed to do when on the highway when an unpredicted storm comes through? Park on the highway when others are moving? the only "exit" would have been through a guard rail off a cliff into a river. Oh, I know - I guess the answer is not be on the road at all from October through May....