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

Tesla autopilot HW3

This site may earn commission on affiliate links.
imagine if Tesla is really able to release a Level 2 system that can handle all driving, including urban driving
c6a523218828fe4286098d9423cf86ad.jpg
 
  • Funny
Reactions: rnortman
Unlike with Lennon, it is hard to do with Tesla.

Maybe this is more like Ariana Grande’s Imagine.

Feels like forever, baby, I never
Thought that it would be you
Tell me your secrets, all of the creep ***
That’s how I know it’s true


Imagine if Tesla had been and would be brutally open and honest about their progress instead?
 
Last edited:
  • Like
  • Funny
Reactions: rnortman and croman
Uhh....what?

Audi’s Level 3 system, Traffic Jam Pilot — which is currently not available to customers — can only do lane keeping and car following at speeds up to 37 miles per hour on divided highways. So that’s less advanced than a Level 2 car that can, say, drive on any public roadway anywhere in the contiguous United States with no driver input. In all traffic conditions and weather conditions. Doing everything that a human driver does.
 
Last edited:
Audi’s Level 3 system, Traffic Jam Pilot — which is not currently not available to customers — can only do lane keeping and car following at speeds up to 37 miles per hour on divided highways. So that’s less advanced than a car that can, say, drive on any public roadway anywhere in the contiguous United States with no driver input. In all traffic conditions and weather conditions.

The tech behind Audi’s Traffic-jam Pilot is capable of doing much more speedwise, it chooses not to because it relieves the driver of responsibility and that raises the bar.

It is thus far more advanced than any system that requires driver to maintain attention because it can handle all obstacles unlike say a Tesla that does not even have obstacle detection for many types of obstacles yet — thus using the driver as a crutch.

It is of course true the Traffic-jam Pilot’s sensor suite is designed for highway driving only.
 
  • Like
Reactions: rnortman
Are we comparing apples to apples here. If Audi has L3 - I don't care how restricted - it's something Tesla doesn't have. And yeah, Tesla might have an L2 feature set that's better and/or broader than Audi has, IDK. But the fact remains: L3 is true autonomy while L2 is not. Because with L3 the system not only monitors, but is responsible for monitoring the environment
 
  • Helpful
  • Like
Reactions: rnortman and OPRCE
Audi’s Level 3 system, Traffic Jam Pilot — which is not currently not available to customers — can only do lane keeping and car following at speeds up to 37 miles per hour on divided highways. So that’s less advanced than a Level 2 car that can, say, drive on any public roadway anywhere in the contiguous United States with no driver input. In all traffic conditions and weather conditions. Doing everything that a human driver does.

That means both Audi's system and Tesla's system (in this theoretical example) are improperly labeled as L3 and L2 respectively.
 
Are we comparing apples to apples here. If Audi has L3 - I don't care how restricted - it's something Tesla doesn't have. And yeah, Tesla might have an L2 feature set that's better and/or broader than Audi has, IDK. But the fact remains: L3 is true autonomy while L2 is not. Because with L3 the system not only monitors, but is responsible for monitoring the environment

Autopilot can already do everything that Traffic Jam Pilot can do. Autopilot just doesn’t give the driver permission to take their hands off the wheel and watch TV.

Is Traffic Jam Pilot more technologically advanced than Autopilot? If so, how so exactly? And how do we know?

Is it more technologically advanced, or has Audi just removed a safeguard? Audi could require the driver to pay attention at all times, and that would most likely be safer.

Taking away a safety measure doesn’t in itself make a product more technologically advanced. And unless there is some independent evidence that a product is actually more technologically advanced, why should we believe that it is?

Audi could sell a car without seatbelts, and say it’s more advanced because it doesn’t require seatbelts. But why doesn’t it require seatbelts? Just because Audi says so? Why not put in seatbelts anyway?

That means both Audi's system and Tesla's system (in this theoretical example) are improperly labeled as L3 and L2 respectively.

The SAE Levels of Automation are whack. Or at least, the implication that higher levels are more advanced is only true if all else is equal. Including which specific driving tasks are performed in which specific environments, and the system’s level of risk tolerance. Otherwise, there are all kinds of cases where a hypothetical Level 3 or 4 system is doing something much easier and more limited than a hypothetical Level 2 system.

LevelsofDrivingAutomation.png


SAE Levels of Driving Automation | Center for Internet and Society
 
Last edited:
Autopilot can already do everything that Traffic Jam Pilot can do. Autopilot just doesn’t give the driver permission to take their hands off the wheel and watch TV.

Because Autopilot can not do what Audi’s system can: detect all classes of obstacles.

It is not only a question of reliability. It is a lack of feature. Autopilot does not identify all classes of obstacles (which is required for vision to detect them) whereas Audi does through the use of Lidar. Tesla simply lacks this feature — and even after Tesla implements this feature, getting its reliability to Lidar level through vision probably needs time. But most importantly Tesla simply lacks this feature in anything we’ve seen so far as many classes of obstacles are not identified.

By the way, Audi does this bit already today in consumer cars, the Lidar ships in eg Audi A8 for its AEB even though Level 3 AI does not yet.

This is mandatory for Level 3+ and a big reason why I do not expect to sleep in my Tesla’s driver’s seat in 2020. If Tesla was using MobilEye EyeQ4 with a much more robust and proven vision track-record then maybe but even MobilEye is looking at Lidar for their autonomous car...
 
Last edited:
  • Like
Reactions: OPRCE
A real example. All of Waymo’s vehicles have safety drivers, so they are Level 2. Nuro is delivering groceries with a lil’ robot that drives on select suburban streets. Nobody is in the robot at any time, so it’s Level 4.

Ergo, Nuro is ahead of Waymo on the SAE scale.


Also, since Waymo’s vehicles are Level 2 and Autopilot is Level 2, Waymo One and Autopilot are the same level of automation according to the SAE scale. Oh, and Audi’s Traffic Jam Pilot is more automated than Waymo One vehicles.
 
A real example. All of Waymo’s vehicles have safety drivers, so they are Level 2. Nuro is delivering groceries with a lil’ robot that drives on select suburban streets. Nobody is in the robot at any time, so it’s Level 4.

Ergo, Nuro is ahead of Waymo on the SAE scale.

Waymo has had customers without safety drivers too, stop repeating a misinformation. Many news stories about this discussed here too. There was even an interview of one such customer who reported paying much more attention on that ride...

The simply choose to have safety drivers at this stage.
 
  • Like
Reactions: OPRCE

Being designed for Level 4 or being technically capable of Level 4 doesn’t mean that a system can safely do Level 4, let alone that it actually does Level 4.

Arguably, if a vehicle can’t be trusted to do driving without a safety driver, then it isn’t technically capable. (Otherwise, what does “technically capable” mean?)

For example, Navigate on Autopilot is a Level 4 system if you just allow it operate with the driver paying attention, or being expected to take over. But that wouldn’t be safe.

So, it seems like Waymo One and Navigate on Autopilot are the same in that respect. Both could be Level 4 if the requirement for a safety driver were removed. But neither can safely remove that requirement, so neither are actually Level 4.
 
The SAE Levels of Automation are whack. Or at least, the implication that higher levels are more advanced is only true if all else is equal. Including which specific driving tasks are performed in which specific environments, and the system’s level of risk tolerance. Otherwise, there are all kinds of cases where a hypothetical Level 3 or 4 system is doing something much easier and more limited than a hypothetical Level 2 system.

What you're saying doesn't make any sense. You don't get to pick and choose what features you want to implement on a per level basis.
 
For example, Navigate on Autopilot is a Level 4 system if you just allow it operate with the driver paying attention, or being expected to take over. But that wouldn’t be safe.

Completely false.

Level 4 must allow for circumstances where the human driver does not pay attention or does not intervene. You can set limits to what circumstances this can happen in (location, weather, speed range etc) but not to this fact itself. That is why it is called ”mind off” (Level 3 is eyes off because there you must intervene after a given time).

In what circumstances can NoA be driven in this manner? None. The manufacturer does not allow for the driver to not pay attention and the driver is expected to take over at all times. Hence it is Level 2 (because it controls more than one parameter, TACC only is Level 1).

With Waymo the limitations are geofenced but it can be done and has been done without an attentive driver — Waymo is Level 4.
 
Last edited:
What you're saying doesn't make any sense. You don't get to pick and choose what features you want to implement on a per level basis.

This is how the SAE Levels of Automation are defined. Brad Templeton has a good post about this: A Critique of NHTSA and SAE "Levels" of self-driving

“In reality, there are not levels, but a set of capabilities and features, which operate on a certain subset of the roads.”
He gives the example of an autonomous campus shuttle (max speed 20 km/h) that is Level 4, versus a Google/Waymo vehicle that is at most Level 3 — actually Level 2, as we now know.
 
This is how the SAE Levels of Automation are defined. Brad Templeton has a good post about this: A Critique of NHTSA and SAE "Levels" of self-driving

“In reality, there are not levels, but a set of capabilities and features, which operate on a certain subset of the roads.”
He gives the example of an autonomous campus shuttle (max speed 20 km/h) that is Level 4, versus a Google/Waymo vehicle that is at most Level 3 — actually Level 2, as we now know.

I honestly can't tell if you're just trolling or you don't understand what the SAE Levels mean.
 
Volvo is planning to launch a feature called Highway Pilot in 2021. (Man, all these companies are so original with their naming...) Highway Pilot, as envisioned, is a Level 4 system, since it lets you sleep while it’s activated. But, as the name suggests, it only works on highways. Per the SAE definition, it handles all aspects of the dynamic driving task, but only in one driving mode: highway driving. That makes it a Level 4 system.

“Volvo’s primary goal “is to get to Level 4 for the unsupervised part of the journey,” R&D chief Henrik Green told reporters at the factory opening here Wednesday.​

By “unsupervised,” Highway Pilot will allow drivers to let the Volvo XC90 drive itself without having to be ready to take over.”

https://www.automobilemag.com/news/spa-gen-2-xc90-coming-2021-offer-level-4-autonomy/

A system by Tesla that did all modes of driving — highway, city street, dirt road, etc. — but required human supervision for all modes would be a Level 2 system. That’s how Level 2 is defined. If there’s constant supervision, it’s Level 2, regardless of the feature set.

So, Highway Pilot would be Level 4, whereas a feature complete Tesla FSD would be Level 2. But Tesla’s FSD would have more capabilities and features in terms of the driving tasks it can actually perform.