Wiki Consumer AV - Status Tracking Thread

[EVNow]

This thread is to track the progress of Consumer AVs - from ADAS to full self-driving. I'll keep this info updated.

Let me know if you have any updates & corrections.

 

[S4WRXTTCS]

It's missing Rivian.

As I understand it the goal is hands free L2 with auto lane change (initiated by driver).

It's also missing Volvo's L3 that they have planned for California. Not sure I'd include a Cali on system. If that gets added than maybe add MB L3 in Germany. Both of those would go in another category as its actually autonomous.

 

[EVNow]

It's missing Rivian.

As I understand it the goal is hands free L2 with auto lane change (initiated by driver).

It's also missing Volvo's L3 that they have planned for California. Not sure I'd include a Cali on system. If that gets added than maybe add MB L3 in Germany. Both of those would go in another category as its actually autonomous.

I'll checkout Rivian & Volvo L3.

MB Drive Pilot is already there ...

What do you mean they would go in another category ? Note that there are several low speed L3 highway products are already in the table. L3 is just "no monitoring".

ps :

Volvo confident it can get its ‘unsupervised’ highway driving mode approved in California

The feature will be available in the upcoming electric SUV.

www.theverge.com

 

[Daniel in SD]

It wasn't good enough to make self-driving meaningless? Now we have to make AV meaningless too?
None of these systems could be categorized as autonomous except Tesla FSD and the various "eyes off" traffic jam pilot systems.
Maybe we should call them "Driving Automation Systems"

 

[EVNow]

It wasn't good enough to make self-driving meaningless? Now we have to make AV meaningless too?
None of these systems could be categorized as autonomous except Tesla FSD and the various "eyes off" traffic jam pilot systems.
Maybe we should call them "Driving Automation Systems"

Full Autonomy has not been achieved - so we are tracking progress towards it.

India's "first war of independence" was in 1857. Actual independence would not be achieved for another 90 years.

 

[S4WRXTTCS]

What do you mean they would go in another category ? Note that there are several low speed L3 highway products are already in the table. L3 is just "no monitoring".

L3 isn't just no monitoring, but the liability is transferred to the car.

It's a world of difference than L2 and is deserving of its own category.

Now keep in mind I do like this idea, and comparing systems.

 

[ThomasD]

What would be nice is level 3 while towing a trailer

 

[EVNow]

L3 isn't just no monitoring, but the liability is transferred to the car.

It's a world of difference than L2 and is deserving of its own category.

I can mention L3 for sure. But we do want to keep them all in the same table for comparison.

Liability transfer is just a product of failure rates. I’d rather they publish failure rates directly for comparison.

 

[Daniel in SD]

I can mention L3 for sure. But we do want to keep them all in the same table for comparison.

Liability transfer is just a product of failure rates. I’d rather they publish failure rates directly for comparison.

Most importantly it's about avoiding a felony conviction!

 

[EVNow]

Most importantly it's about avoiding a felony conviction!

Hope you are not speaking from experience

Seriously, here is the question about L3 in consumer cars.

Do Merc & BMW really have production L3 in Germany ? How did they test & confirm it works ? How many miles did their testers drive ?

I think thats going to be an issue ... esp. if they are looking for < 60 km/h conditions and millions of miles of driving.

Same question for Volvo. How can they assure error rates are good if they can't test a million miles ?

The only videos of Merc I can find of L3 are the demo ones in Merc's demo track where they actually recreate slow spped traffic-jam situation.

 

[Daniel in SD]

Hope you are not speaking from experience

No, but it is something I'd like to avoid!

Seriously, here is the question about L3 in consumer cars.

Do Merc & BMW really have production L3 in Germany ? How did they test & confirm it works ? How many miles did their testers drive ?

I think thats going to be an issue ... esp. if they are looking for < 60 km/h conditions and millions of miles of driving.

Same question for Volvo. How can they assure error rates are good if they can't test a million miles ?

I have no idea. I suspect that the systems are actually very primitive and they're just going to trust LIDAR and a "don't hit anything" algorithm. The braking distance from 60kph is less than 60ft. It will be interesting to see how it works out.
Testing for a million miles doesn't seem that expensive. $50 an hour / 20 mph * 1e6 = $2.5 million = chump change. I do worry that you need a lot more distance than that to really find edge cases.

 

[EVNow]

I have no idea. I suspect that the systems are actually very primitive and they're just going to trust LIDAR and a "don't hit anything" algorithm. The braking distance from 60kph is less than 60ft. It will be interesting to see how it works out.
Testing for a million miles doesn't seem that expensive. $50 an hour / 20 mph * 1e6 = $2.5 million = chump change. I do worry that you need a lot more distance than that to really find edge cases.

Lets do some simple economics math.

When Merc sells a car and adds an L3 feature - say for $5k, they make $1k Margin. Lets say over the life of the car it travels 20k miles on highways < 60 kmph. A crash would mean $50k costs (car + any injuries). So, at a minimum you want to make sure you have no more than 1 accident per 50 cars i.e. 1 crash in 50*20k = 1M miles. To get any kind of confidence on crash rates, Merc needs to run for 10 Million miles and get less than 10 crashes.

To run 10 Million miles is the problem. Not because of the money it takes to pay test drivers, but because of time. How many cars would they need to run for how many months to get 10 Million miles on highways at less than 37 mph / 60 kmph ? I know CA has a lot of traffic jams - still, you can't probably get more than 6 hours a day of traffic jams < 37 mph or about 250 miles per car per day. So, they need to run for 400 days using 100 cars to get to 10 Million miles. Its the limited ODD that is the issue.

 

[EVNow]

Here is an independent 3rd party youtube video of Nio Autopilot (they call it Pilot).

Nio Navigate on Pilot (NOP) is like NOA but doesn't yet have all the functionality of NOA - but has more than just lane keeping. It can overtake and can get on ramp & offramp.

 

[Daniel in SD]

Lets do some simple economics math.

When Merc sells a car and adds an L3 feature - say for $5k, they make $1k Margin. Lets say over the life of the car it travels 20k miles on highways < 60 kmph. A crash would mean $50k costs (car + any injuries). So, at a minimum you want to make sure you have no more than 1 accident per 50 cars i.e. 1 crash in 50*20k = 1M miles. To get any kind of confidence on crash rates, Merc needs to run for 10 Million miles and get less than 10 crashes.

To run 10 Million miles is the problem. Not because of the money it takes to pay test drivers, but because of time. How many cars would they need to run for how many months to get 10 Million miles on highways at less than 37 mph / 60 kmph ? I know CA has a lot of traffic jams - still, you can't probably get more than 6 hours a day of traffic jams < 37 mph or about 250 miles per car per day. So, they need to run for 400 days using 100 cars to get to 10 Million miles. Its the limited ODD that is the issue.

Yeah, I doubt they're going to test even close to 10 million miles. Maybe they'll try to extrapolate from close calls?
I don't see how the limited ODD makes that big a difference. Of course your system would be very optimized for the 405 over Sepulveda pass. haha. So you've got to have 4x as many test vehicles because you can't test 24/7?

 

[EVNow]

So you've got to have 4x as many test vehicles because you can't test 24/7?

Yes - I think that is the main issue. Even my 6 hours per day is probably very optimistic. There will be lot of times even within those 6 hours when the speeds are higher than 37 mph, so L3 won't work. Or the speeds are much less than 37 mph, so the car will travel less miles ! So, 4x is optimistic.

BTW, if they find issues, they need to correct and retest !

This is where Tesla's approach of 60k fleet and ADAS until they can get to L3 shines. They can keep running it as L2 until the disengagement rates suggest L3 is ok.

ps : For Volvo, won't they have to test in CA - and won't they have to register to test autonomous driving and file reports ?

 

[Daniel in SD]

Yes - I think that is the main issue. Even my 6 hours per day is probably very optimistic. There will be lot of times even within those 6 hours when the speeds are higher than 37 mph, so L3 won't work. Or the speeds are much less than 37 mph, so the car will travel less miles ! So, 4x is optimistic.

BTW, if they find issues, they need to correct and retest !

This is where Tesla's approach of 60k fleet and ADAS until they can get to L3 shines. They can keep running it as L2 until the disengagement rates suggest L3 is ok.

Mercedes could of course do the same thing since they also have an L2 mode.
There's a reason no one has released an "eyes off" highway system, it's a lot harder than most people think!
Tesla has literally billions of miles of data from divided highways and they won't do it either. I guess it is true that robotaxis are infinitely more valuable than any highway system, even one without the low speed restriction.

 

[EVNow]

There's a reason no one has released an "eyes off" highway system, it's a lot harder than most people think!

Some of the things I read on Merc/BMW in Germany suggested the problem is German regulation. The regulation restricts L3 to 60 kmph. I guess they probably have the same issues I'm posting it here - how do the OEMs demonstrate L3 is safe for regulatory approval.

If I was a regulator, I'd ask for exactly what I wrote above - show test drives worth atleast 10x the distance to failure. Those are extremely time consuming for tiny fleets.

ps :

Is 2020 The Year For Eyes-Off Automated Driving?

When can we start driving a vehicle equipped with a Level 3 eye-off automation? Anyone not looking at the road now is being reckless. Automakers are moving this new capability to the market fast–but how fast?

www.forbes.com

For example, take Audi’s bold pronouncement that the 2018 Audi A8 would offer a Level 3 Traffic Jam Pilot. They did so, and although vehicles could be purchased with this feature in Germany, once on the road lack of clarity with German road regulations forced Audi to refrain from enabling it. Ouch.​

​

Audi gives up on Level 3 autonomous driver-assist system in A8

Audi says it no longer plans to offer Traffic Jam Pilot in the current generation of the A8.

www.motorauthority.com

 

[EVNow]

Here is an interesting document :

UN Regulation No. 157 - Automated Lane Keeping Systems (ALKS)​

UN Regulation No. 157 - Automated Lane Keeping Systems (ALKS) | UNECE

unece.org

Page 59 onwards list what tests need to be conducted. Has zero references to anything with failure rates rates or how many miles of testing are needed. Its more like a one day test showing various features work (in ideal conditions, I guess). Infact it has gems like this ...

The test shall be executed at least:
(a) With a minimum test duration of 5 minutes;

 

[Bladerskb]

This thread is to track the progress of Consumer AVs - from ADAS to full self-driving. I'll keep this info updated.

As you can see there are a lot of questions here I couldn't find answers to. Let me know if you can find links with answers to these questions. Also, any updates & corrections.

View attachment 761503


View attachment 761504

good list, one correction though is that Ultra Cruise is point to point and availablility is through OTA in 2022
Also i think you should break this further in ADAS (L2) and ADS (Automated Driving System, L3,L4...)

Consumer AV Status - Highway
Consumer AV Status - City Streets
Consumer ADAS Status - Highway
Consumer ADAS Status - City Streets

Also the Toyota Drive is on-ramp to off-ramp
Also honda traffic jam pilot use Mobileye's EyeQ4 based vision.
There are definitely more changes that could be made.

 

[EVNow]

good list, one correction though is that Ultra Cruise is point to point and availablility is through OTA in 2022

So, like FSD Beta ?

Also the Toyota Drive is on-ramp to off-ramp

But does it make needed change from one freeway to another ? Can't find any actual reviews or videos in the wild. Only thing I found was an early video that said the Drive worked only in some segments of the freeway - let alone on-ramp to off-ramp.

ps : BTW, when going from one freeway to another, is it considered off-ramping ?