Autonomous Car Progress

[Bladerskb]

blah blah blah.

All I here is a bunch of Tesla fanatics who yet again defines any and everything as "Is this what Tesla is doing? If yes: Good If no: bad".
Even things that were "good" are now considered "Bad" by the same people. Autonomous L4 cars are no longer autonomous because other companies actually have them. Because Tesla doesn't have it, L4 is now no different than L2. Just like with "HD map", if Tesla ever gets to the point that they can deploy a driverless fleet in a locale (city/state) and they utilize remote operation (which they said in 2019 that they will). These same fantastic will be back here and when asked about how remote op means the car isn't autonomous....

 

[Bitdepth]

The SYNOPSIS image at the bottom of your post makes my point exactly: Level 5, FULL AUTOMATION, says "The vehicle performs all driving tasks under all conditions. Zero human attention or interaction is required.

Level 4 can require human input at times (though it must be able to park itself safely first) but Full Automation, what I've been calling Full Autonomy, by definition, never requires any human interaction. Though as before, I'd make an exception for plugging in to the charger, or filling the tank if the car is a stinker.

From Level 2 to Level 4 is huge. Level 2 is Supervised Self-Driving, while Level 4 is Self-Driving with the option to request assistance after stopping in a safe place. Level 2 exists now and works wonderfully. Level 5 seems out of reach in the near term, IMO. Level 4 seems to be everybody's goal for now. Not "Full Automation" but "High Automation."

The difference between L4 and L5 is limited ODD. L4 must perform all driving task without human intervention under its ODD. Human intervention is not required. But just because a L5 fully automated vehicle fails at some road somewhere does not make it not fully automated. Fully automated as stated in J3016 and totally autonomous are not the same thing. A vehicle operating within its ODD by definition is totally autonomous within its operational design domain. To circle back to remote assist as covered by J3016 does not make a vehicle not highly or fully automated. It is just a matter of fact of how these things will operate in the real world. Same way planes have airtraffic control is the same way L4 and L5 vehicles operated in a fleet will have remote assist and fleet management.

Pg14.
Providing fleet asset management services to vehicles while in-use (e.g., managing emergencies, summoning or providing remote assistance as needed, responding to customer requests and break-downs).

That synopsis image states

L4 "the vehicle performs all driving tasks under specific circumstance. Geofencing is required, human override is still an OPTION"
L5 "the vehicle performs all driving tasks under all conditions. Zero human attention is required."

This is what happens when you jump in the middle of a discussion without reading the context. That is why i asked what does not fully autonomous or not autonomous means.

Not totally autonomous. By history.

Check out the experience in AZ. The traffic cone debacle illustrated both remote and roving support driver needed to free it. Not autonomous.

Read the 100 pages of posts. And so when it fails, it's no longer autonomous. Like L2, but with fewer interventions. And should there even be "severe" cases in a carefully circumscribed L4 geofence? Shouldn't it be able to handle construction zones flawlessly?

Paying passengers will not be amused to be late to their destination because the Waymo refused to proceed when it gets confused.

Your posts suggest Waymo fanboi tendencies. Is diplomat your cousin?

The least we can do as enthusiast is start from a common base of understanding what we are talking about from the source material where all ADS engineers and laws are based on. Waymo fanboy tendencies. Lol

 

[stopcrazypp]

Of course the quality of service matters. None of know how often Waymo rides fail to complete or how many vehicles can be serviced by a single remote assistant. Hopefully we'll get a better picture soon when these services deploy in San Francisco. I'm guessing there will be much more customer and media scrutiny there.

I think Waymo needs to prove themselves to their parent Alphabet more. Alphabet have suddenly pulled projects that run too long without commercialization, so I think that'll push Waymo to quicken their pace.

I'm not sure however how much the public will get a view in the operation. Passengers are still under NDA.

 

[Daniel in SD]

I think Waymo needs to prove themselves to their parent Alphabet more. Alphabet have suddenly pulled projects that run too long without commercialization, so I think that'll push Waymo to quicken their pace.

I'm not sure however how much the public will get a view in the operation. Passengers are still under NDA.

I wonder how much Alphabet is funding Waymo anymore (they probably know better. haha)? They've been getting money from VCs.
The Waymo One service in Chandler doesn't have an NDA so if there is one in SF hopefully it won't last long.

 

[rxlawdude]

The least we can do as enthusiast is start from a common base of understanding what we are talking about from the source material where all ADS engineers and laws are based on. Waymo fanboy tendencies. Lol

Agreed. In law, the court will first look to the "plain meaning" of a word (e.g., dictionaries, legal dictionaries), then move on to ascertaining legislative/regulatory intent as well as the interpretation of the word in prior cases.

We can agree that the intent of Tesla is to get to what is known as "L5." We can agree the intent of Waymo, as stated, is to be "L4."

Can we agree that, given what beta 10 FSD does, that Tesla is already "autonomous" but with continuous human supervision? See, that's all I'm expecting in the coming months. Anything more than that were operative aspirations.

However, for that 2017 M3, I do expect any hardware and CPU upgrades to be performed by Tesla in order to approach those aspirations.

For our 2020 MY, my expectations are at least for NOA on virtually any public road, city or highway. I have little expectation of hardware upgrades to get to full(er) autonomy.

 

[rxlawdude]

Why should anyone care if a vehicle is truly "autonomous" or not if they can legally sleep in the backseat or watch a movie?

Again, the SAE doesn't use the word "autonomous" anyway.

Because this will require a sea change in liability and insurance law. This assumes the manufacturer will indemnify the sleeping passenger in the "driver's" seat.

We're in for interesting times with respect to these changes, on top of the technology we love to discuss here.

 

[stopcrazypp]

I wonder how much Alphabet is funding Waymo anymore (they probably know better. haha)? They've been getting money from VCs.

Estimate from here is 92.5% owned by Alphabet on last $2.25 billion funding round (based on $30 billion valuation) and then 86-89% for latest $4 billion based on $50-$100 billion valuation. It largely depends on valuation (and how much of the funding round was non-Alphabet), but whatever numbers you use, it's a very long way away from Alphabet losing their large controlling stake.
Alphabet Stock: Making The Right Moves With Waymo (NASDAQ:GOOG)

The Waymo One service in Chandler doesn't have an NDA so if there is one in SF hopefully it won't last long.

The Chandler program looks like launched in April 2017 and they lifted their NDA at almost exactly 2 years later (April 2019), although people were still not allowed to talk about their experiences on their rides before the lifting of NDA.
Waymo launches self-driving car service Waymo One – TechCrunch

https://www.reddit.com/r/SelfDrivingCars/comments/biz4fb

Yeah, hopefully it's lifted a lot sooner for the SF NDA (which Waymo did confirm exists).

https://twitter.com/x/status/1430221217001218048

 

[Daniel in SD]

Because this will require a sea change in liability and insurance law. This assumes the manufacturer will indemnify the sleeping passenger in the "driver's" seat.

We're in for interesting times with respect to these changes, on top of the technology we love to discuss here.

Waymo has had plenty of collisions. As far as I know no passenger has been held liable. It seems like no change is required.
For all the upcoming L3 systems the manufacturer is liable for collisions.

Estimate from here is 92.5% owned by Alphabet on last $2.25 billion funding round (based on $30 billion valuation) and then 86-89% for latest $4 billion based on $50-$100 billion valuation. It largely depends on valuation (and how much of the funding round was non-Alphabet), but whatever numbers you use, it's a very long way away from Alphabet losing their large controlling stake.

Of course but as long as they can keep getting funding from VCs I don't see why Alphabet would shut it down? At least if they keep raising money there is some chance of getting a return on their investment. If they liquidate they'll get very little. Obviously the calculus changes if they can no longer get funding and they need to put their own money in.

 

[rxlawdude]

Waymo has had plenty of collisions. As far as I know no passenger has been held liable. It seems like no change is required.
For all the upcoming L3 systems the manufacturer is liable for collisions.

Of course but as long as they can keep getting funding from VCs I don't see why Alphabet would shut it down? At least if they keep raising money there is some chance of getting a return on their investment. If they liquidate they'll get very little. Obviously the calculus changes if they can no longer get funding and they need to put their own money in.

Waymo Robotaxi != Tesla used by owner.

Where can I find an enumerated list of "plenty of [Waymo] collisions?"

 

[Daniel in SD]

Waymo Robotaxi != Tesla used by owner.

Where can I find an enumerated list of "plenty of [Waymo] collisions?"

Elon has already said Tesla will be like the Otis Elevator Company, they will be liable for defects in the system.
Waymo safety white paper and California autonomous collision reports (Autonomous Vehicle Collision Reports - California DMV)

 

[Terminator857]

Elon has already said Tesla will be like the Otis Elevator Company, they will be liable for defects in the system.
Waymo safety white paper and California autonomous collision reports (Autonomous Vehicle Collision Reports - California DMV)

Many or most? of those collisions are Waymo getting rear ended while Waymo is stopped. I'm aware of one collision that was obviously Waymo's fault. A few were fault isn't obvious, but most are other parties fault. Do we know how many are Waymo's fault? Any of those more than a minor fender bender?

 

[FloridaJohn]

Can we agree that, given what beta 10 FSD does, that Tesla is already "autonomous" but with continuous human supervision?

Human supervision and intervention.

You would also have to look at the disengagement rate as well. I wouldn’t consider something “autonomous” if it required an intervention every couple of miles.

 

[Daniel in SD]

Many or most? of those collisions are Waymo getting rear ended while Waymo is stopped. I'm aware of one collision that was obviously Waymo's fault. A few were fault isn't obvious, but most are other parties fault. Do we know how many are Waymo's fault? Any of those more than a minor fender bender?

No, we don't know anything about fault. That's not really relevant to safety though since the goal is reduce the number and severity of collisions regardless of fault. Waymo vehicles do get rear ended a lot, probably because their behavior is unpredictable to human drivers.

 

[rxlawdude]

That's not really relevant to safety though since the goal is reduce the number and severity of collisions regardless of fault. Waymo vehicles do get rear ended a lot, probably because their behavior is unpredictable to human drivers.

That will be a neat trick. As I mentioned, AI anticipating unpredictable biologic units' actions is the real challenge.

 

[Terminator857]

... Waymo vehicles do get rear ended a lot, probably because their behavior is unpredictable to human drivers.

As Waymo becomes more common I suppose it will become more predictable. I wonder how many accidents have happened that were not caused by Waymo not being predictable and instead caused by driver inattentiveness. I suspect most of recent accidents fall in that category. We can't expect a driverless car to be in zero accidents until humans stop driving.

 

[daniel]

The difference between L4 and L5 is limited ODD. L4 must perform all driving task without human intervention under its ODD. Human intervention is not required. But just because a L5 fully automated vehicle fails at some road somewhere does not make it not fully automated. Fully automated as stated in J3016 and totally autonomous are not the same thing. A vehicle operating within its ODD by definition is totally autonomous within its operational design domain. To circle back to remote assist as covered by J3016 does not make a vehicle not highly or fully automated. It is just a matter of fact of how these things will operate in the real world. Same way planes have airtraffic control is the same way L4 and L5 vehicles operated in a fleet will have remote assist and fleet management.


That synopsis image states

L4 "the vehicle performs all driving tasks under specific circumstance. Geofencing is required, human override is still an OPTION"
L5 "the vehicle performs all driving tasks under all conditions. Zero human attention is required."

This is what happens when you jump in the middle of a discussion without reading the context. That is why i asked what does not fully autonomous or not autonomous means.

Again you make my point. Thank you. L5: "Zero human attention is required." L4: Human intervention is allowed.

I doubt I'll see L5 in my lifetime. I begin to lose hope of being able to buy a car capable of Level 3 operation in the city where I live.

L4 can be anything from 20 sq mi in Chandler, to almost L5 but needs the occasional assist. The SAE level definitions certainly have their usefulness, but when a car company announces that they have a Level 4 car that tells me nothing of whether it will do anything at all for me. This is why for us ordinary people who want to buy a driverless car, different definitions entirely are needed.

 

[johnm]

"L4 can be anything from 20 sq mi in Chandler" ... it can be substantially less. Here is a shuttle that drives an 18 block route:

Med City Mover - MnDOT

The Minnesota Department of Transportation is continually striving to improve, meet customer needs, get the most from every dollar, and be transparent about our progress and use of public funds.

www.dot.state.mn.us

I assume the vehicle is:

EZ10 passenger shuttle | EasyMile

easymile.com

 

[Bladerskb]

Really nice video from Pony.Ai on how their autonomous stack handles construction zones. I especially like that the video takes you through step by step what the perception sees and what the planning does.

But but HD maps are useless because what if the hd map and perception disagree?

 

[daniel]

But but HD maps are useless because what if the hd map and perception disagree?

Lots of times our senses disagree, and our brain resolves the disagreement. Not always correctly, but it doesn't mean our senses are useless. HD maps are not useless. The software just needs to be able to resolve disagreements in a way that contributes to the overall system being safer than a human driver.

The problem with HD maps is not that they might sometimes disagree with perception. The problem is that they are expensive to produce, and therefore are unlikely to be available over large areas any time in the near to medium future. They can serve well for robotaxis that only operate over a small area. But consumer cars intended to be sold nation-wide or world-wide will need to be able to operate without them.

 

[Bladerskb]

Lots of times our senses disagree, and our brain resolves the disagreement. Not always correctly, but it doesn't mean our senses are useless. HD maps are not useless. The software just needs to be able to resolve disagreements in a way that contributes to the overall system being safer than a human driver.

The problem with HD maps is not that they might sometimes disagree with perception. The problem is that they are expensive to produce, and therefore are unlikely to be available over large areas any time in the near to medium future. They can serve well for robotaxis that only operate over a small area. But consumer cars intended to be sold nation-wide or world-wide will need to be able to operate without them.

This problem has already been solved by Mobileye's REM. They are available TODAY. Worldwide. Over 625 million miles of roads already. Not sure why this keeps coming up. This is what Mobileye Supervision runs on. The only con right now is increasing the refresh rate. They need to get it down to hourly refresh rate. They were at quarterly update in early 2020. Maybe they are at monthly or weekly today. I don't know, but its the only axis left for them to improve.