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

Can level 5 Autonomy be achieved with Hardware suite 2.0?

This site may earn commission on affiliate links.
I didn't see writtin, or hear spoken, "hinging on regulatory approval" and cetainly not "but we're not sure if the hardware suite is going to actually be able to handle it".
Here's what it says on the order page (bolding is by Tesla):
"Please note that Self-Driving functionality is dependent upon extensive software validation and regulatory approval, which may vary widely by jurisdiction. It is not possible to know exactly when each element of the functionality described above will be available, as this is highly dependent on local regulatory approval. Please note also that using a self-driving Tesla for car sharing and ride hailing for friends and family is fine, but doing so for revenue purposes will only be permissible on the Tesla Network, details of which will be released next year."

Worst case scenario if for legal or software reasons Tesla's suite can't offer self driving, they refund you the $3000 for the option (or whatever it is for your country).
 
  • Like
Reactions: Johan
Here's what it says on the order page (bolding is by Tesla):
"Please note that Self-Driving functionality is dependent upon extensive software validation and regulatory approval, which may vary widely by jurisdiction. It is not possible to know exactly when each element of the functionality described above will be available, as this is highly dependent on local regulatory approval. Please note also that using a self-driving Tesla for car sharing and ride hailing for friends and family is fine, but doing so for revenue purposes will only be permissible on the Tesla Network, details of which will be released next year."

So written then. Mental note taken, thanks!
 
What Tesla describes seems not to be vague! You can start your driverless ride hailing service via Tesla Network.

So L5 is real.
L5 hasn't been defined yet, we don't know what limitations are acceptable until the public weighs in.

What I understand is: L5 does not mean that all accidents, injuries and deaths will be zero.

Autonomous Driverless cars can reduce them but not eliminate them.
That's a red herring. For example, you can avoid an accident by having a car needlessly block traffic. And that may be preferable. But it's a failed system as far as substituting for a human.
 
L5 hasn't been defined yet, we don't know what limitations are acceptable until the public weighs in.

That's a red herring. For example, you can avoid an accident by having a car needlessly block traffic. And that may be preferable. But it's a failed system as far as substituting for a human.
The media keep saying level 5, but it seems Elon only mentioned that was a possibility with the new hardware (not a hard promise). Tesla in written text promises "self-driving" which may not necessarily be level 5 (however you want to define level 5).
 
The media keep saying level 5, but it seems Elon only mentioned that was a possibility with the new hardware (not a hard promise). Tesla in written text promises "self-driving" which may not necessarily be level 5 (however you want to define level 5).

There seems to be a movement away from the levels because everyone seems to be at odds with the transition. Level 3, and Level 4 have transitions from the car driving to requiring the human driver. Where L5 doesn't have that messiness, and I view Full Self-Driving as L5 driving. You did leave off the FULL from the self-driving.
 
The media keep saying level 5, but it seems Elon only mentioned that was a possibility with the new hardware (not a hard promise). Tesla in written text promises "self-driving" which may not necessarily be level 5 (however you want to define level 5).
May not necessarily be even "level 4", and he's also only "thinks" that it will be available in 2018. But that's their roadmap and the implication. I wonder how the car will find a numbered parking spot? Can it read numbers directly below it on the ground? Do you manually train it to park in the correct spot?
 
...That's a red herring....

I am just getting ready for disappointment for a system that I thought that it should be foolproof but it may not meet my expectation after all:

Self-Driving Cars Could Cut Down on Accidents, Study Says

It says it can avoid 90% of accidents but to me, the failure rate of 10% is quite high of a rate!

This article says the same rate:

Will Self-Driving Vehicles Eliminate The Need For Auto Insurance?

The accident avoidance rate is lower according to Foxx, United States Secretary of Transportation, NHTSA:

Foxx said: “It’s not a surprise that at some point there would be a crash of any technology that’s on the road....

Foxx said that he expected the autonomous technology not to be perfect, but that it could prevent up to 80% of crashes in the US....
 
Last edited:
May not necessarily be even "level 4", and he's also only "thinks" that it will be available in 2018. But that's their roadmap and the implication. I wonder how the car will find a numbered parking spot? Can it read numbers directly below it on the ground? Do you manually train it to park in the correct spot?

We should have a thread of all the edge cases and what if's for a Full Self-Driving Tesla.
 
I am just getting ready for disappointment for a system I thought that it should be foolproof but it may not meet my expectation:

Self-Driving Cars Could Cut Down on Accidents, Study Says

It says it can avoid 90% of accidents but to me, the failure rate of 10% is quite high of a rate!

This article says the same rate:

Will Self-Driving Vehicles Eliminate The Need For Auto Insurance?

The accident avoidance rate is lower according to Foxx, United States Secretary of Transportation, NHTSA:

Foxx said: “It’s not a surprise that at some point there would be a crash of any technology that’s on the road....

Foxx said that he expected the autonomous technology not to be perfect, but that it could prevent up to 80% of crashes in the US....

You have some unrealistic expectations, don't you?

Let me give you a corollary to help you understand why it will be obvious that we should accept a non-perfect autonomous technology that is much safer than manual driving: When computers started handling accounts in banks in stead of manually kept ledgers with human doing manual calculations it was clear that the error rate became much lower, but certainly not zero. But that doesn't mean that people long to go back to the old way of doing things that was even more prone to error!
 
...But that doesn't mean that people long to go back to the old way of doing things that was even more prone to error!

After watching all the science fiction of driverless cars and all the publicity about the new technology, I was in disbelief and shocked when I read that driverless technology can still fail 10 to 20%!

That's crazy! I told myself. But I have gradually accepted it.

It's just real life and you just have to do the best and keep improving!

Thanks for the example @Johan! It does make sense.
 
Last edited:
There seems to be a movement away from the levels because everyone seems to be at odds with the transition. Level 3, and Level 4 have transitions from the car driving to requiring the human driver.
The definition of level 4 does not require any hand off. It is just that the car can't operate in certain conditions (presumably it'll come to a safe stop if it detects such conditions during the course of travel). A Level 3 car is the one that requires hand off.

Where L5 doesn't have that messiness, and I view Full Self-Driving as L5 driving. You did leave off the FULL from the self-driving.

Even if you keep the "full" part, there is still a level of messiness because the SAE definition changes two parameters at once moving from L4 to L5 and is still a bit vague.

Level 4: driving-mode performance vs Level 5: full performance
Level 5: all roadway and environmental conditions that can be managed by a human driver

Driving mode is a type of driving scenario with characteristic dynamic driving task requirements (e.g., expressway merging, high speed cruising, low speed traffic jam, closed-campus operations, etc.).
http://www.sae.org/misc/pdfs/automated_driving.pdf

This makes in unclear in colloquial usage. Wired counts the google car as a "level 5", however, it doesn't meet the strict definition of the SAE article above (particularly it can't handle highway or bad weather conditions, although it tells you that it doesn't before it starts the journey and presumably it can come to a safe stop on the side of the road if conditions deteriorate).
Everyone Wants a Level 5 Self-Driving Car—Here’s What That Means

Here's another definition that says level 5 is a car that has no option for human driving (no steering wheel or controls). Obviously the Model S won't meet that definition.
Autonomous driving levels 0 to 5: Understanding the differences - TechRepublic

Another interpretation can be level 5 is a car that can operate with no one inside (even if it doesn't meet all driving modes/weather conditions).

Basically, if a car was like the Google car (which requires no driver and even has no controls), but can't operate in certain driving modes (like on the highway) is it still "level 5"? Can it be considered "full self-driving"?

And if it can handle all driving modes, but not necessarily all weather conditions, is it still "level 5"? Can it be considered "full self-driving"?

What if it can handle all driving modes, all weather conditions, but there are still some exceptions where it doesn't operate like a human (like the policeman directing traffic example people like to bring up). Is it "level 5"? Is it "full self-driving"?

Under the current vague/contradictory definitions, it seems open to interpretation.

I'm not even sure if Elon said level 5 specifically or if reporters interpreted that full self-driving = level 5?
 
Last edited:
FWIW, I doubt that Tesla AP2.0 hardware would pass regulator requirements that today's autonomous cars pass in Singapore.
I do think that Tesla is using Lidar in company vehicles to train their Lidarless AP2.0 vehicles.
If push came to shove, adding a $150 Lidar to a lidarless (but lidar trained) Tesla AP2.0 would require minimal validation issues and be an effective workaround. Perhps not viable for USA cars

Last Thursday, due to a road accident ahead, I was required to do a U turn (which is normal in my state, but technically illegal) along a road outside of a hospital. there were 2 ambulances there, but the police had not arrived. I followed the direction of citizens traffic control (plain clothed nurses were directing the traffic)
could driverless cars handle this, No.
could autopilot car handle this, yes with driver temporarily taking over control of the vehicle.
 
It's not that vague:
http://www.sae.org/misc/pdfs/automated_driving.pdf as posted by stopcrazypp

Keep in mind:
Dynamic driving task includes the operational (steering, braking, accelerating, monitoring the vehicle and roadway) and tactical (responding to events, determining when to change lanes, turn, use signals, etc.) aspects of the driving task, but not the strategic (determining destinations and waypoints) aspect of the driving task.

LevelsofDrivingAutomation.png
 
Location of the cameras means, that there are some edge cases, where car can't manage. As you know, side facing cameras are in B-pillars and wide lens camera in front 3 cameras set up has 120 degrees view. So that means, that if B-pillars are blocked, car doesn't have visibility to sides. See this case

https://www.google.com/maps/@60.165...4!1sHMKwGUVkYOL-4MLyqTmdRg!2e0!7i13312!8i6656

Imagine, that you are coming to the street from that very tight corridor from inner court. Buss parked blocks visibility to the left. This is very difficult place for human driver also, but I guess human would very slowly creep and stretch his head forward to see if there is a car coming from left (there is!). But car can't stretch his head and as side facing camera to the left is blocked by bus and wide angle camere in front has 120 degrees vision, car does not have any visibility to the left.

But I guess one could say, that this case is not safely manageable by human either,
 
Last edited:
Location of the cameras means, that there are some edge cases, where car can't manage. As you know, side facing cameras are in B-pillars and wide lens camera in front 3 cameras set up has 120 degrees view. So that means, that if B-pillars are blocked, car doesn't have visibility to sides. See this case

https://www.google.com/maps/@60.165...4!1sHMKwGUVkYOL-4MLyqTmdRg!2e0!7i13312!8i6656

Imagine, that you are coming to the street from that very tight corridor from inner court. Buss parked blocks visibility to the left. This is very difficult place for human driver also, but I guess human would very slowly creep and stretch his head forward to see if there is a car coming from left (there is!). But car can't stretch his head and as side facing camera to the left is blocked by bus and wide angle camere in front has 120 degrees vision, car does not have any visibility to the left.

But I guess one could say, that this case is not safely manageable by human either,

You realize there are four side cameras right... (and they overlap viewing angles)
 
  • Like
Reactions: MP3Mike
Imagine, that you are coming to the street from that very tight corridor from inner court.
A Tesla wouldn't fit inside that hallway. Problem solved. :)

But seriously, I think there are bound to be some edge cases where a single radar, multiple camera system will struggle. For example, it's afternoon, and you're traveling East on the highway. The sun is low in the sky, behind you. The car needs to change lanes but it's rear-facing cameras are blinded by the low sun and the ultrasonics are only short-range. Can it detect a rapidly approaching car from the back?

The argument would be that humans use only vision to navigate, not radar, but human vision is also in stereo, providing depth, and, in such sunlit scenarios, we also use sunglasses. )

There had been speculation about radar in all four corners, which, at a high level, sounds like it would have been a more costly, but perhaps a more robust solution for edge cases.

EDIT: I'm also curious if any of the forward cameras are color. How can the system tell the difference between a white line and a yellow line, for example? I'm thinking of the demonstration video where the car crossed a center white line, but also parking situations where the curb may be painted.
 
  • Like
Reactions: William3