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

Mercedes-Benz receives world's first internationally valid system approval for conditionally automated driving

This site may earn commission on affiliate links.
Since we have zero safety data we're both using our personal common sense to predict how confusing it will be.
It has nothing to do with "safety data".

It a UX question. Let us say I'm sitting in slow moving traffic on a busy freeway. If it is at 36 mph, Merc is responsible. If its at 38 mph, I'm responsible.

If I'm the lead on the project - I'd not approve of such a system. If you think designing and delivering such a product is kosher, good luck with that.
 
  • Disagree
Reactions: ItsNotAboutTheMoney
We've had conversations before about L3 and the view that it could be a not-very-useful middle ground, autonomous while it's going (yay!) but throwing control back to the driver too often or too quickly. The generally-agreed observation was that the market would decide how valuable this would be - and that it wouldn't prove very valuable if any competitor had something better, that handed off much less often.

This seems to me to be exactly the situation with the MB system here.

I'll admit that there may be a small subset of people with long commutes in heavy rush hour on freeways, that stay below 37 mph for long periods. But it doesn't seem to me to be the basis of a major milestone in the History of Self-Driving Technology.

Volvo is promising something similar, but I think the intent is to support a more practical range of highway speeds, basically an entrance-to-upcoming-exit ODD. I'm sure MB will get there too before long, I just think they've jumped the gun in order to claim First.

Tesla is constantly criticized, somewhat justifiably, for falling short of Elon's ambitious goals and timelines for FSD. And I've been very critical of their overly limited hardware suite. But at least they are ambitious goals. I think that all of those giving kudos to MB for this highly limited "achievement" would be mercilessly pounding Tesla if they introduced exactly the same L3 feature with exactly the same limitations.
 
Effectively less than my 2015 MS 70 with Autopilot hardware version 1

However, with the more modern Model 3, it still failed the 2019 AAA Emergency Braking Test at 30 MPH, which is below Mercedes L3 speed of 37 MPH.


Model 3 also failed again in 2020 China in front of the public and press:




This is why Tesla is refusing to classify its Autopilot and FSD as L3 and above.
 
Last edited:
It has nothing to do with "safety data".

It a UX question. Let us say I'm sitting in slow moving traffic on a busy freeway. If it is at 36 mph, Merc is responsible. If its at 38 mph, I'm responsible.

If I'm the lead on the project - I'd not approve of such a system. If you think designing and delivering such a product is kosher, good luck with that.
Do we know exactly what the UX is?
I doubt what you describe would even be allowed. I'd bet the system will use traffic data to give the driver plenty of warning and it will not oscillate between modes.
To me safety is the most important question with any automation system. And of course that includes the safety of the transitioning between modes.
 
  • Like
Reactions: momo3605
I think that all of those giving kudos to MB for this highly limited "achievement" would be mercilessly pounding Tesla if they introduced exactly the same L3 feature with exactly the same limitations.
I doubt it. Many Tesla owners have been asking for "eyes off" highway driving even if it's only at lower speeds.
Why are they working on city streets when they have billions of miles of NoA data already? I think a L3 system that worked everywhere NoA currently works would sell a lot of cars.
 
Do we know exactly what the UX is?
I doubt what you describe would even be allowed. I'd bet the system will use traffic data to give the driver plenty of warning and it will not oscillate between modes.
To me safety is the most important question with any automation system. And of course that includes the safety of the transitioning between modes.
This is my belief also. If one is supposed to sleep, work or do anything without paying attention, a sudden handover is not acceptable - and I guess Mercedes has thought about that. But we shall see when cars hit the roads next year.

Others though see an opportunity to spread some FUD about the competitor, maybe to avoid questions on why Tesla did not go that route. Maybe Tesla can deliver this next year also? They should try.
 
I doubt it. Many Tesla owners have been asking for "eyes off" highway driving even if it's only at lower speeds.
Why are they working on city streets when they have billions of miles of NoA data already? I think a L3 system that worked everywhere NoA currently works would sell a lot of cars.
OK, we'll have to have our different conclusions. I simply can't see the critics nor many of the enthusiasts being happy with freeway "eyes off / hands off" that calls you back every time the traffic jam moves at 37 MPH. And I can't see this as being the breakthrough event to trigger a new paradigm of insurance coverage wiih associated data logging and fault adjudication.
 
  • Like
Reactions: stopcrazypp
That definitely applies to you !

I hope Tesla doesn’t do this BS 37 mph limit on freeways.

What nonsense.

From time to time, Tesla Autopilot/FSD just suddenly hands back its control to me with no warnings at all at all speed, slow or fast.

It just gives me a loud alert noise and a red steering icon and all controls becomes manual with zero second transition time.

The advantage of L3 is it gives you a transition time so you can take over and start to supervise the L2 ADAS shortly.

It can gives you warnings ahead because it can predict when the traffic will exceed 37 MPH shortly based on the movement of several cars ahead.

The car directly in front is stationary but 5 or more cars beyond that started to move. It then calculates when the 4th car in front would move and so on...
 
  • Like
Reactions: ItsNotAboutTheMoney
From time to time, Tesla Autopilot/FSD just suddenly hands back its control to me with no warnings at all at all speed, slow or fast.

It just gives me a loud alert noise and a red steering icon and all controls becomes manual with zero second transition time.

The advantage of L3 is it gives you a transition time so you can take over and start to supervise the L2 ADAS shortly.

It can gives you warnings ahead because it can predict when the traffic will exceed 37 MPH shortly based on the movement of several cars ahead.

The car directly in front is stationary but 5 or more cars beyond that started to move. It then calculates when the 4th car in front would move and so on...

Exactly.

I'm glad I don't have to deal with regular slow-moving highway traffic.
I've often read here about how Autopilot allows people to relax physically in heavy traffic. This is an extension where people could relax mentally in what can be a tedious and stressful part of driving.

If traffic is switching frequently between speeds above and below the threshold, then drivers will stick to Level 2, but where traffic is crawling, they'll let L3 take over.
 
This is my belief also. If one is supposed to sleep, work or do anything without paying attention, a sudden handover is not acceptable - and I guess Mercedes has thought about that. But we shall see when cars hit the roads next year.

Others though see an opportunity to spread some FUD about the competitor, maybe to avoid questions on why Tesla did not go that route. Maybe Tesla can deliver this next year also? They should try.
Note you are not allowed to sleep in a SAE L3 car. You must always be ready to take over shortly after it requests you to do so. Sleeping is not compatible with that. While you don't have to pay full attention (like L2), you still need to be in a state you can take over in seconds.
For SAE the requirement is "several seconds":
SAE J3016 User Guide
For the UN R157e standard that Mercedes is following (in section 5.4.4.1 as above), it's 10 seconds.
https://unece.org/sites/default/files/2021-03/R157e.pdf
 
I doubt it. Many Tesla owners have been asking for "eyes off" highway driving even if it's only at lower speeds.
Why are they working on city streets when they have billions of miles of NoA data already? I think a L3 system that worked everywhere NoA currently works would sell a lot of cars.
Tesla owners are calling for eyes off at the same capability as NOA or even AP generally (meaning full speed, and in a variety of roads). I doubt many people are asking for a system that was subject to the same limitations as the Mercedes system (37 mph traffic jams in mapped divided highways designed for much higher speeds).
 
Tesla owners are calling for eyes off at the same capability as NOA or even AP generally (meaning full speed, and in a variety of roads). I doubt many people are asking for a system that was subject to the same limitations as the Mercedes system (37 mph traffic jams in mapped divided highways designed for much higher speeds).
I think with a little refinement and effort, Tesla COULD turn basic AP or even NoA into an L3 system. At least a lot less effort than L5 FSD or whatever the goal is right now.

That being said, if regulations only allow for 37mph L3, I’d happily take it as a stepping stone to L3 full speed AP.
 
That being said, if regulations only allow for 37mph L3, I’d happily take it as a stepping stone to L3 full speed AP.
I thought it was Mercedes-Benz that set the 37MPH (60kPH) boundary for this particular system. I certainly hope that isn't written into any regulations; it's inappropriate, possibly corrupt and classically erroneous for government regulators to apply quantitative targets and/or mandated engineering solutions to a rapidly-developing technology.
 
I thought it was Mercedes-Benz that set the 37MPH (60kPH) boundary for this particular system. I certainly hope that isn't written into any regulations; it's inappropriate, possibly corrupt and classically erroneous for government regulators to apply quantitative targets and/or mandated engineering solutions to a rapidly-developing technology.
How many systems out there do you think would take real responsibility for the driving task at 60MPH+ in 2021/2022? That's going to be another huge step considering how reaction times change at higher speed, and frankly I'd need to have a lot of trust in a system to stop paying attention to the road when traveling at highway speeds. Things can go wrong real quick at 37MPH much less 60-70MPH, and the chances of dying in an accident increase with the speed at which things can go wrong.

I don't think there's a chance in hell that Tesla would do this with Autopilot right now, you need to be monitoring Autopilot / FSD at all times
 
  • Funny
Reactions: jtsai1031
so apparently we are just going to trust that Mercedes' ADAS system is somehow orders of magnitude more reliable in some way over Tesla? let's be honest here. these tactics smell of obvious desperation. they are going to find ways to market their tech as superior without having to prove as such. hey look the gov't says our cars can drive themselves without supervision (so implicitly they are better right??) but is that really the case?? has the reliability really been shown to be that much better?? or are they just taking a gamble before Tesla drinks all their milkshake?
 
honestly don't even think unsupervised FSD will ever be allowed if humans are still allowed to drive. i could be wrong and i hope i am but the problem really seems to be that human drivers wouldn't be able to cooperate with robots. just look at how those grocery store bots are treated lol. wait until humans start exploiting Autonomous Vehicles. only way this works is if humans are prohibited from driving altogether.
 
I don't think it's a massive assumption that all of these companies have autonomous systems capable of attempting complex maneuvers when vehicles were attempting them back in 2016, the biggest distinction here is that Tesla allows their consumer vehicles to attempt them while calling the systems Level 2 and making the driver responsible. The other companies are generally far more conservative in their approach and lean more towards offering simpler systems with narrower functions aiming for higher SAE levels.

Mercedes is partnered with NVIDIA and uses their Drive system, you can find heavily edited and curated videos of it doing FSD Beta-like stuff


Now I'm not quite suggesting NVIDIA's / Mercedes' system would do as well in these maneuvers, although who knows what point each system is at, but I have no doubt all of them are capable of attempting similar stuff. And I think you'd find capabilities across the systems is much closer than it appears in the public realm.

I personally don't care much about what a vehicle will attempt if I need to supervise it, but I'll give money to whoever gets Level 3+ autonomy and sells it at a price that makes sense for the value it'll add by allowing me to focus on other stuff as it drives in its defined ODD. If Tesla puts out a Level 3+ highway system that will do my 10hr highway drives while I work or surf the web or whatever, I'm there. If Mercedes does it, I'm there.
but i think you are confusing the generalized solution Tesla is working on with a more limited system like Waymo (or in this case MB). from all I have seen Tesla is tackling the much harder but more necessary solution AND they have apparently the best engineers (Karpathy etc) working on it. last i heard Mercedes had given up on a full autonomous solution. isn't this 37 mph bs just a marketing gimmick?
 
but i think you are confusing the generalized solution Tesla is working on with a more limited system like Waymo (or in this case MB). from all I have seen Tesla is tackling the much harder but more necessary solution AND they have apparently the best engineers (Karpathy etc) working on it. last i heard Mercedes had given up on a full autonomous solution. isn't this 37 mph bs just a marketing gimmick?
I don't think that point is lost on me, I've said many times in this thread that Tesla is taking a fundamentally different approach to this. True autonomy, to the extent where we could have a consumer vehicle that doesn't require a steering wheel or pedals, is far more difficult and I think is off in the relatively distant future. It makes me wonder why Tesla isn't pursuing some of the steps between the current state and something that I don't think will be realized for many years to come.

If you think that goal will be hit sooner, then Tesla's approach seems like a no-brainer. But when I think of all the situations that a car would need to handle to actually drive you around in all situations without supervision, which would be required to sell a consumer vehicle with full autonomy, I don't think we're even close. And I personally won't pay for a system that requires my constant supervision regardless of what it will or won't attempt and possibly successfully execute.
 
Note you are not allowed to sleep in a SAE L3 car. You must always be ready to take over shortly after it requests you to do so. Sleeping is not compatible with that. While you don't have to pay full attention (like L2), you still need to be in a state you can take over in seconds.
For SAE the requirement is "several seconds":
SAE J3016 User Guide
For the UN R157e standard that Mercedes is following (in section 5.4.4.1 as above), it's 10 seconds.
https://unece.org/sites/default/files/2021-03/R157e.pdf
Yep, correct, I just read it. UN states the transition as at least 10 seconds. There are also requirements for driver to be in seat and have eyes open, and surveillance of that.

That makes the value quite a bit lower. How great a 20 min nap each afternoon on the way home would be. Almost like on a train!