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

Mercedes approved for ACTUAL self driving in the USA. And will accept responsibility.

This site may earn commission on affiliate links.
I don't see L3 highway being viable other than in small ODD like low speed traffic jam. I just think L3 at full highway speeds is too risky. Imagine L3 where the human is allowed to read a book or watch a movie, and then something sudden happens that the L3 can't handle. There won't be enough time for the human to regain situational awareness and resume control. And a situation where the human was reading a book and then the L3 tells it take over but the human fails to take over in time and the car crashes at 70 mph, would be disastrous. So L3 highway would need to be able to handle all those split second safety critical issues and fallback on its own, just like L4. So I think L3 at full highway speeds would need to be close to L4 highway anyway. L3 is really only good for ODD where there is enough time to hand control back to the human. That's why L3 traffic jam makes sense since it is low speed, low risk, with enough to give control back to the human. Alternatively, you could use driver monitoring to make sure the human does not read or watch a movie, to make sure they are able to resume control pretty quickly. That would help make sure that the human is ready to take control if need be. But that would sort of defeat the benefit of L3, that the human can read a book or watch a movie.
This is just the first slow-speed release. It is working on the next high-speed 80MPH release:


L3 restrictions will still be there, but the speed will be higher.

If the car senses that the current 80MPH speed will be stopped on the freeway because of an accident in front, it will hand it over to the driver. If the driver doesn't accept, it will stop itself from 80MPH.

mqdefault.jpg


A LIDAR can see as far as 500 meters (1,640 feet) away, so there's still time to alert the driver to take over.
 
Last edited:
  • Helpful
Reactions: diplomat33
This is just the first slow-speed release. It is working on the next high-speed 80MPH release:


L3 restrictions will still be there, but the speed will be higher.

If the car senses that the current 80MPH speed will be stopped on the freeway because of an accident in front, it will hand it over to the driver. If the driver doesn't accept, it will stop itself from 80MPH.

mqdefault.jpg


A LIDAR can see as far as 500 meters away, so there's still time to alert the driver to take over.

Thanks. I know about the speed increase. I remain a bit skeptical of L3. But having lidar that can see 500 m away, will be very helpful. That should give the human enough time to take over. I guess I just prefer L4 where you remove the human element altogether and just make the car good enough on its own. That seems like a better approach to me than letting the human not pay attention but then rely on them being able to resume control in x seconds.
 
  • Like
Reactions: Tam
...I guess I just prefer L4 where you remove the human element altogether and just make the car good enough on its own. That seems like a better approach to me than letting the human not pay attention but then rely on them being able to resume control in x seconds.
No question, but what are we doing while waiting for L4?

Letting customers buy something like L3 is better than skipping L3 and waiting for consumer L4 in the meantime.

Consumer L4 is complex. Thus, by imposing lots of restrictions, as in L3, it's a good temporary bridge to L4.
 
  • Like
Reactions: APotatoGod
No question, but what are we doing while waiting for L4?

Letting customers buy something like L3 is better than skipping L3 and waiting for consumer L4 in the meantime.

Consumer L4 is complex. Thus, by imposing lots of restrictions, as in L3, it's a good temporary bridge to L4.

Sure. Like I said, L3 is meant to be a bridge to L4. I am optimistic about consumer L4. I believe consumer L4 is ~5 years away. If true then I think "hands off L2" can be the bridge, we don't need L3. If consumer L4 is 10 years away as Spacecoin believes, then yeah, I would agree we need L3 as a bridge. So I guess it depends how quickly you believe consumer L4 will happen.
 
I disagree, my interpretation is that L3 will be as capable and safe as L4 in the same ODD. But, the L3 car has a driver that can be handed control when the car nears the boundaries of the ODD.
The L4 though must manage all by itself, all the time. If a L4 car sense an outside ODD state, it must stop. Like a stuck Cruise.

If Cruise was L3, it would have a sleeping bio-backup in the car, and wake him up to solve the situation.
 
  • Like
Reactions: spacecoin
I disagree, my interpretation is that L3 will be as capable and safe as L4 in the same ODD. But, the L3 car has a driver that can be handed control when the car nears the boundaries of the ODD.
The L4 though must manage all by itself, all the time. If a L4 car sense an outside ODD state, it must stop. Like a stuck Cruise.

No, L4 can also hand control back to the human when it nears the boundaries of the ODD.
 



“Mercedes is the first automaker to receive such approval in the U.S., beating out names like Tesla, General Motors, Ford, and even Honda”
According to the Mercedes web site, Drive Pilot is an extremely limited system that only operates under the following conditions:

  • Clear lane markings on approved freeways
  • Moderate to heavy traffic with speeds under 40 MPH
  • Daytime lighting and clear weather
  • Driver visible by camera located above driver's display
  • There is no construction zone present.
So it doesn’t have to worry about pedestrians, bicyclists, cross traffic, traffic lights, stop signs, turn lanes, et cetera so on and such forth.

If you frequently find yourself in traffic jams on divided highways, you’ll love it. But it’s hardly comparable to FSD.
 
  • Like
Reactions: diplomat33
Correct. That's my point. True L3 highway would be very risky because of the issue of giving the human enough time to resume control and making sure the human is able to resume control in time and would basically need to be very close to L4 anyway for safety issues. Or you force the human to be alert and ready to take over at a moments notice and then you don't have true L3 anymore. So in both cases, L3 is not really viable IMO unless you are doing a very small, low risk ODD like traffic jam scenarios.



Yes.
So, few people will buy Mercedes Drive Pilot?
 
No, L4 can also hand control back to the human when it nears the boundaries of the ODD.
There is really no difference between an L3 and and L4 given the same ODD and reliability given that definition.

Imagine a sudden hailstorm or fog or any other condition that’s outside of the ODD. That will need an awake fallback ready driver, no? That would make it an L3 for me unless it can handle all weather and all road situations, which seems unlikely for a personally owned vehicle in the coming 10 years.

Sometimes it snows in April, as Prince sang :)
 
Last edited:
There is really no difference between an L3 and and L4 given the same ODD and reliability given that definition.

No. The main difference is the fallback. L3 requires a human fallback, L4 does not.

Here is precisely what the SAE levels say about this:

L3

DDT fallback-ready user (while the ADS is engaged):
• Is receptive to a request to intervene and responds by performing DDT fallback in a timely manner
• Is receptive to DDT performance relevant system failures in vehicle systems and, upon occurrence, performs the DDT fallback in a timely manner
• Determines whether and how to achieve a minimal risk condition
• Becomes the driver upon disengagement of the ADS

L4

Passenger/dispatcher (while the ADS is engaged):
• Need not perform the DDT or DDT fallback
• Need not determine whether and how to achieve a minimal risk condition
• May perform the DDT after the ADS reaches its ODD limit (see 5.5 NOTE 2)
• May request that the ADS disengage
• May become the driver after a requested disengagement
 
Last edited:
I disagree, my interpretation is that L3 will be as capable and safe as L4 in the same ODD. But, the L3 car has a driver that can be handed control when the car nears the boundaries of the ODD.
The L4 though must manage all by itself, all the time. If a L4 car sense an outside ODD state, it must stop. Like a stuck Cruise.
An L4 car can hand the control to a driver also if it exits the ODD, the main difference is if the driver does not respond, it knows how to reach a safe state.
If Cruise was L3, it would have a sleeping bio-backup in the car, and wake him up to solve the situation.
NO! You can not have a sleeping driver in an L3 vehicle. By definition, the driver must be ready to take over in several seconds. Sleeping is not compatible with this!
 
  • Like
Reactions: APotatoGod
No. The main difference is the fallback. L3 requires a human fallback, L4 does not.

Here is precisely what the SAE levels say about this:

L3

DDT fallback-ready user (while the ADS is engaged):
• Is receptive to a request to intervene and responds by performing DDT fallback in a timely manner
• Is receptive to DDT performance relevant system failures in vehicle systems and, upon occurrence, performs the DDT fallback in a timely manner
• Determines whether and how to achieve a minimal risk condition
• Becomes the driver upon disengagement of the ADS

L4

Passenger/dispatcher (while the ADS is engaged):
• Need not perform the DDT or DDT fallback
• Need not determine whether and how to achieve a minimal risk condition
• May perform the DDT after the ADS reaches its ODD limit (see 5.5 NOTE 2)
• May request that the ADS disengage
• May become the driver after a requested disengagement
Yes. You’re making my point. The ODD limit can be hit instantly without prior warning by traffic conditions or weather which effectively requires a fallback ready driver? It can’t really MRM on a limited access highway if there is a heavy smoke on the roads.

By when to you expect personally owned L4 on the highway that can handle all weather and traffic situations either by being awesome or teleops so that the driver need not to be fallback ready and can sleep? Seems at least 10 years away to me at this point.

It also seems to me you’d be able to hit L3 a lot sooner, if the OEM:s want to leave the cozy L2 “driver is always responsible“ because you can limit the L3 ODD to a fraction of what the L4 ODD would be?

Finally. in order to build the L4 with fallback that you are proposing you need to build L3 DDT fallback, so why not release that first in a more limited ODD (which it enables) to get some revenue?
 
Last edited:
but someone has to have done it by now ! I can't find any footage, is it really available or vaporware?
You can buy the DrivePilot L3 package in Germany. "Fahrassistenz-Paket mit DRIVE PILOT 8.841,70 €" (about 10k USD). The L2 alone is 2-3k.

 
but someone has to have done it by now ! I can't find any footage, is it really available or vaporware?
I haven't heard any delivery for this 2024 L3 system in the US just yet.

However, there have been many journalists who tried it out on the road.


It's been sold in Germany for over a year with no accidents during its L3 operation.

The no accidents part could be explained because of all the restrictions: slow speed, daylight, no city streets, no auto lane change (that’s L2 mode so Mercedes doesn't have to pay in an L2 lane change collision)...
 
  • Like
Reactions: spacecoin
Not cheap. $2,500/year. If I keep my L3 for 8 years, that's $20,000 down the hole.
But if you choose to sell your car or it is totaled in a crash, you only have paid for the feature while you were using it.

And you don't have to pay upfront for effectively nothing but a poorly implemented L2, which is what Tesla purchasers of FSD did. I didn't understand when we made our purchase in 2020. Having heard the hype, I thought it would be L3 within a couple of years. Now I don't believe it will be L3 within the 8 years I usually keep a vehicle and if my car is totaled or I choose to move to a different tesla with better hardware to support L3/4 (not effin likely but a possible scenario for others), my original never delivered FSD is not transferable to another tesla.

So, to me, this expensive option is for something that exists and probably represents recovery of the tech development. Unlike tesla which had me pay towards the tech development but never actually delivers.
 
It's all part of the process.
They use previous versions to test the new ones.
If you want to skip L3, then yeah, it'll take way longer to get L4 acceptably "ready".
But if you release a limited L3 first, gather some data, improve it slowly, release a less restricted version of L3, gather more data... so on and so forth.
That's a much quicker way (scientific method) to get to a usable L4.

It might take 15 yrs to get to L4 if you try skipping L3. But only 5 if you don't.
The reality is, at this point, the hardware exists that's good enough for L4 (will still keep improving though).
It's the software that takes so long. Coding for every possible scenario takes a long time. So they have to limit it to what they have coded.
AI can help, but it's still new and needs to be quadruple checked.
But it's all part of the progression process. It cannot, and should not, be rushed.
 
Yes. You’re making my point. The ODD limit can be hit instantly without prior warning by traffic conditions or weather which effectively requires a fallback ready driver? It can’t really MRM on a limited access highway if there is a heavy smoke on the roads.

If the ODD limit can be hit instantly without prior warning then it cannot work for L3. With L3, the human might be reading or watching a movie but needs to be the fall-back ready user when prompted. So the human will need some time to stop what they are doing, regain situational awareness and resume control. That only works in situations where the L3 can give the human sufficient advance notice. So L3 only makes sense in an ODD where there is enough time for the human to react. So any ODD limit that can happen without warning cannot be an ODD limit for L3.

L3 must be able to handle everything in the ODD that happens quickly and must only need to ask the human to take over when it can give sufficient advance notice. Also remember that L3 relies on the human to do the fallback. If the human cannot do the fallback, L3 will likely be in trouble. That is why L3 requires robust driver monitoring to make absolutely sure that the human is able to do the fallback.

So the manufacturer who wants to do L3 will need to design their L3 and select the proper ODD to make sure that the L3 can safely handle everything inside the ODD and that the human can safely take over every time with enough advance notice.

By when to you expect personally owned L4 on the highway that can handle all weather and traffic situations either by being awesome or teleops so that the driver need not to be fallback ready and can sleep? Seems at least 10 years away to me at this point.

Teleops at highway speeds is too risky IMO. So L4 highway will simply need to be awesome, as you put it. I think that is a big reason why Waymo is not doing driverless on highways yet, despite testing on highways for over a year now. Waymo knows that the Waymo Driver needs to be able to safely handle all highway scenarios without any remote assistance at all, ie it needs to be awesome.

I do expect L4 highway that can handle all traffic situations in good weather in a few years. L4 highway in all weather might be 10 years away.

It also seems to me you’d be able to hit L3 a lot sooner, if the OEM:s want to leave the cozy L2 “driver is always responsible“ because you can limit the L3 ODD to a fraction of what the L4 ODD would be?

Yes, you can do L3 in a more limited ODD, like traffic jam assist. That is what I am saying. But if you want to do a bigger ODD, like all highways from on ramp to off ramp, all weather, at full highway speeds, then I think you basically need L4 so that I think doing L4 highway makes more sense.

Finally. in order to build the L4 with fallback that you are proposing you need to build L3 DDT fallback, so why not release that first in a more limited ODD (which it enables) to get some revenue?

Perhaps the reference to L4 asking the human to take over when leaving the ODD confused you because it sounds like L3. The key difference is that with L4, human fallback is optional but with L3, human fallback is required.

Example: L4 highway approaches highway exit. It asks the human to take over. If human takes over, L4 is off, and human resumes driving. If human fails to take over, L4 does the fallback. It pulls over as soon as it is safe to do so. L3 approaches the highway exit. It requests the human take over. The human must take over since the L3 cannot do the fallback. See the difference?

So you don't need to do L3 human fallback in order to do L4. You can build L4 to do its own fallback. L4 does not require human fallback. Human fallback is optional in L4.
 
Last edited:
I haven't heard any delivery for this 2024 L3 system in the US just yet.

However, there have been many journalists who tried it out on the road.


It's been sold in Germany for over a year with no accidents during its L3 operation.

The no accidents part could be explained because of all the restrictions: slow speed, daylight, no city streets, no auto lane change (that’s L2 mode so Mercedes doesn't have to pay in an L2 lane change collision)...
If it's sold , where are the videos ? those with journalists and safety drivers does not count