Autonomous Car Progress

[Bladerskb]

It's the same platform they used in AZ for ride hailing which the I-Pace inherited a lot of things from. I think it's fair to leave out vehicles like the Firefly perhaps, but the Pacificas were designed for ride-hailing just like the I-Pace.

That's not a fair account either. Waymo has been testing extensively using the Pacificas since before then in SF. In 2020 they removed them from SF for a short while due to possible unrest:
Waymo pulls self-driving cars in San Francisco as cities nationwide brace for unrest

I have seen the Pacificas personally in SF during the pandemic.

The CA disengagement reports say for Waymo:
2014 53,428 miles (unknown model only Sept-December 2014 numbers)
2015 409,759 miles (unknown model)
2016 654,627 miles (unknown model)
2017 294,929 miles (unknown model, although note Pacificas were introduced this year)
2018 Pacificas 1,271,587 miles, I-paces 0 miles
2019 Pacificas 1,454,137 miles, I-paces 0 miles
2020 Pacificas 628,442 miles, I-paces 396.3 miles
2021 Pacificas 302,220 miles, I-paces 2,023,624 miles.

Leaving out 3+million miles the Pacificas tested on Bay Area and SF roads is misleading if you are trying to answer the question of how long it might take to deploy in a "new" city.

Here's the original question for context, person asking obviously already knew that Waymo had extensive experience in SF (so not surprising to him at all for it to take 6 months after ride hailing launch) and was asking about how long it would take for different cities which Waymo have never operated in:

That's not SF, that's majority mountain view. You do realize that CA isn't SF right?
Local reports before Waymo large scale testing was week(s) without seeing a waymo car. After their large scale testing began (late 2020 and fullly realized early 2021) there are reports of people seeing a waymo car every few mins.

TLDR: Before Waymo's large scale testing in SF, spotting a waymo in SF was a rarity, after...spotting a Waymo was common every few minutes.

https://twitter.com/x/status/989909879975378944

https://twitter.com/x/status/1316817655525859328

https://twitter.com/x/status/1491803705133322258

https://twitter.com/jcreech3/status/1477205862393389058

"To give you a sense of how much they have been testing in SF of late, it is now unusual if on a 30-minute bike ride I do not pass at least two
@Waymo vans. They are everywhere."

https://twitter.com/factory1/status/1343089356731338752

"Are there actually way mo WayMo cars on the streets since SF has been a ghost town, or am I just noticing them more? I swear I see one go by my place 3-4 times a day."

https://twitter.com/marcelemoran/status/1323660262725971969

"It used to be that every bike ride through SF would include seeing 1-2 Cruise self-driving cars on the road. Now it's 1-2 Waymo minivans. Lots of testing citywide."

https://twitter.com/marcelemoran/status/1297535175526498304

“There are some days where it can be up to 50. It’s literally every five minutes. And we’re all working from home, so this is what we hear,” King noted.

Dead-End SF Street Plagued With Confused Waymo Cars Trying To Turn Around 'Every 5 Minutes'

A normally quiet neighborhood in San Francisco is buzzing about a sudden explosion of traffic. Neighbors say their Richmond District dead-end street has suddenly become crowded with Waymo vehicles.

sanfrancisco.cbslocal.com

For example waymo has been testing in 25 cities. To give you a glimpse of what that testing looks like, they have 3 cars in NYC.
Now do you consider 3 cars full scale testing? Do you even see the difference between 3 cars testing versus 100 cars testing around the clock?

On a broader note, Its funny that these nuances exist for Tesla but not for others. They must be PERFECT and immaculate.

Cruise launch in SF at night to be careful and want to progressively extend to day time? Response of Tesla fans is: THATS NONSENSE THEY ARE LYING. THEIR SDC SYSTEM JUST DOESN'T WORK. ITS A GIMMICK!!!!

Elon launch FSD Beta and limits it to 100 tesla super fans for almost 1 year and his reasoning is to be careful and want to progressively extend it to more people?
Tesla's fan response: Oh that makes perfect sense. Tesla wants to be safe, to do otherwise is nonsense.

https://twitter.com/x/status/1315616137464307718

https://twitter.com/x/status/1366951771810062336

For others, it need to be L5 that works everywhere in all conditions perfectly or its considered GARBAGE!
For tesla, they can have a completely trash and dangerous system that meets none of the standards above and its still considered AMAZING!

The constant double standard tesla fans generally employ on a daily basis is simply nothing short of amazing!

 

[Daniel in SD]

This is why I don't think we'll ever see an L3 system capable of highway speeds.

Humans absolutely will fall asleep, or they'll get so caught up in whatever they're doing that they don't realize the car is beeping at them.

I think we'll see enough crashes of L3 systems at traffic assist speeds that the systems don't progress further than that.

How often would it happen and how many collisions would it cause? People can and do fall asleep while using Autopilot but I've never heard of a collision from the car stopping in the middle of the road (which is what it does if you continually ignore warnings). All these L3 systems have camera based monitoring to detect if the driver is sleeping too.

 

[Daniel in SD]

With level 3 how many seconds do I have to take over? How many seconds is it until an accident or incident becomes my fault and not the manufactures fault?

UN standard is 10 seconds but it also can't just hand over 1 second away from a brick wall. Practically speaking the vehicle has to be able to handle all emergency situations. How often do emergency situations last more than 10 seconds?

 

[Daniel in SD]

We talked about this before, so no need to revive the topic:
Autonomous Car Progress

In comments to others, you actually agreed with my point that "wide release" will likely be "autosteer in City streets" (AKA door-to-door L2) so not sure why you are pretending that "robotaxis" will be what FSD Beta will be when "wide release" happens.
Poll: FSD Beta V9.1, how close is it to release (i.e. out of beta)?

"Robotaxis" are a lot farther off than wide release of FSD Beta.

I don't want to drive this off into another tangent, but to the original point, my point is it is conceivable for a door-to-door L2 feature like "Autosteer in City Streets" to work in cities that it was never tested in before.

Go back and read it again. What I said is that FSD Beta is the beta of FSD and they will never release it widely. My prediction is that they will release something ore limited in functionality than FSD Beta as "autosteer on city streets." Obviously I could be wrong! But not about FSD Beta, it's the beta version of FSD.

There are no plans yet for anything beyond that at the moment.

It seems like you and @EVNow have some insider info on the new Master Plan...
Though even if Elon tweeted out that FSD Beta is the beta version of FSD I don't think you two would believe it.

 

[Knightshade]

Go back and read it again. What I said is that FSD Beta is the beta of FSD and they will never release it widely. My prediction is that they will release something ore limited in functionality than FSD Beta as "autosteer on city streets." Obviously I could be wrong! But not about FSD Beta, it's the beta version of FSD

Except you are, in fact, wrong. And have been corrected on it repeatedly.

The CA DMV documents make it clear you're wrong on this, in Teslas own words.

FSDBeta is city streets at L2

it is not an early beta of full L5 FSD, and neither capable, or intended of ever being capable of that.

It seems like you and @EVNow have some insider info on the new Master Plan...
Though even if Elon tweeted out that FSD Beta is the beta version of FSD I don't think you two would believe it.

Naah, it's public info thanks to the CA DMV emails being public.

Again- FSDBeta is not "a beta of the final FSD"

It's city streets (their own words)

it's explicitly, only ever, L2 (their own words)

It lacks an OEDR to EVER GO ABOVE L2 AND THEY HAVE NO INTENT ON CHANGING THAT (their own words).

They plan to develop some FUTURE product that WILL offer >L2, but FSDBeta is not it (their own words).


This isn't a debate, you're factually wrong.

 

[Bladerskb]

 

[Daniel in SD]

Except you are, in fact, wrong. And have been corrected on it repeatedly.

The CA DMV documents make it clear you're wrong on this, in Teslas own words.

FSDBeta is city streets at L2

it is not an early beta of full L5 FSD, and neither capable, or intended of ever being capable of that.



Naah, it's public info thanks to the CA DMV emails being public.

Again- FSDBeta is not "a beta of the final FSD"

It's city streets (their own words)

it's explicitly, only ever, L2 (their own words)

It lacks an OEDR to EVER GO ABOVE L2 AND THEY HAVE NO INTENT ON CHANGING THAT (their own words).

They plan to develop some FUTURE product that WILL offer >L2, but FSDBeta is not it (their own words).


This isn't a debate, you're factually wrong.

I've never denied that they told the CA DMV that, I just think they're saying that in order to avoid regulation. I'm certainly curious how they're going to get to millions of miles between interventions without an OEDR capable of going beyond L2.

 

[EVNow]

I've never denied that they told the CA DMV that, I just think they're saying that in order to avoid regulation. I'm certainly curious how they're going to get to millions of miles between interventions without an OEDR capable of going beyond L2.

By ignoring idiotic levels altogether ?

 

[Knightshade]

I've never denied that they told the CA DMV that, I just think they're saying that in order to avoid regulation.

I mean, city streets is the internal name the actual software code had on this since well before the CA DMV stuff (green wrote about this back in at least 2019), and there's nothing IN THE CODE to support an OEDR capable of >L2... so there's no reason to doubt what Tesla told CA was true- nor would it make sense for them to perjure themselves to a government agency in ways easily checked.

Basically every known fact says you're wrong, but you "feel" you must be right anyway.

I'm certainly curious how they're going to get to millions of miles between interventions without an OEDR capable of going beyond L2.

They're going to eventually have new software, that is not the city streets code being tested as FSDBeta, with a more advanced OEDR. Again they explicitly state this in the CA DMV stuff.

It's unclear at this time if that will require next-gen hardware, but it seems increasingly likely.

If that's "just" HW4, or if it'll also include the next-gen camera suite, remains to be seen.

 

[EVNow]

It seems like you and @EVNow have some insider info on the new Master Plan...
Though even if Elon tweeted out that FSD Beta is the beta version of FSD I don't think you two would believe it.

Ofcourse, all of us have the insider info since the master plans are "secret", according to Musk. Why, don't you believe in every word he says ?

Basically every known fact says you're wrong, but you "feel" you must be right anyway.

@Daniel in SD has made it his life's "goal" to prevent others from being able to use FSD Beta. He has written about it explicitly. So, he wants to prove that Tesla is "illegally testing autonomous vehicles" and to get CA DMV to stop them doing it by making FSD Beta illegal.

 

[bradtem]

People should remember that the levels are not a real thing. They were dreamed up by NHTSA regulators very early in the game around 2012, then taken over by SAE which is mostly folks from the car OEM engineering world, not people on the leading self-driving teams. Most of the serious contenders actually building self-driving systems think the levels are stupid and even a bit evil because of how they mislead the public about a technology that is just barely deployed, and certainly wasn't when the levels were made up.

I write a bit more about this in my column this week on the Mercedes offering -- Mercedes Drive Pilot Lets Driver Play Video Games In Traffic Jams, Is It A Big Deal?

While there are many arguments about what it means to have a standby driver (the thing most similar to "level" 3) generally most don't want the driver to sleep, though some want to handle it if they do because, guess what, even if you tell them not to sleep, they still do. (Hell they even fall asleep with Tesla autopilot.) Some think the "level" is more about legal liability than about technical definitions, but it's not settled.

Most, except for Elon Musk, think that ADAS and self-driving are two different things, and you don't just make self-driving by improving ADAS until it's super good. Former Tesla Autopilot lead Sterling Anderson described that approach as trying to build a ladder to the moon.

The idea of levels suggested that it was just a progression of levels. I convinced SAE to put words in the document saying that they were not, and but even so people ignore them.

 

[S4WRXTTCS]

but nobody with FSD Beta would ever confuse FSD Beta for robotaxi

Nobody with at least half a brain would confuse FSD Beta for a L2 door to door system either.

 

[daktari]

People should remember that the levels are not a real thing. They were dreamed up by NHTSA regulators very early in the game around 2012, then taken over by SAE which is mostly folks from the car OEM engineering world, not people on the leading self-driving teams. Most of the serious contenders actually building self-driving systems think the levels are stupid and even a bit evil because of how they mislead the public about a technology that is just barely deployed, and certainly wasn't when the levels were made up.

I write a bit more about this in my column this week on the Mercedes offering -- Mercedes Drive Pilot Lets Driver Play Video Games In Traffic Jams, Is It A Big Deal?

While there are many arguments about what it means to have a standby driver (the thing most similar to "level" 3) generally most don't want the driver to sleep, though some want to handle it if they do because, guess what, even if you tell them not to sleep, they still do. (Hell they even fall asleep with Tesla autopilot.) Some think the "level" is more about legal liability than about technical definitions, but it's not settled.

Most, except for Elon Musk, think that ADAS and self-driving are two different things, and you don't just make self-driving by improving ADAS until it's super good. Former Tesla Autopilot lead Sterling Anderson described that approach as trying to build a ladder to the moon.

The idea of levels suggested that it was just a progression of levels. I convinced SAE to put words in the document saying that they were not, and but even so people ignore them.

I think

Nope, you simply don't understand the core difference between L2 and L3. In L2, you must be monitoring the vehicle at all times. In L3, you don't, but you are still required to respond within seconds when the car requests you to. I suggest you look at the actual SAE definitions and the explanation I posted.

Here's relatively simple chart:
L0-L2 "You must constantly supervise these support features: you must steer, brake, or accelerate as needed to maintain safety"
L3 "When the feature requests, you must drive"
SAE Levels of Driving Automation™ Refined for Clarity and International Audience

Only L4 says: "These automated driving features will not require you to take over driving". This is the standard required to make sleeping safe, given as others mentioned, the sounds the car may make may not be enough to wake you up quickly enough for you to be ready to take over in seconds.

Nope, you simply don't understand the core difference between L2 and L3. In L2, you must be monitoring the vehicle at all times. In L3, you don't, but you are still required to respond within seconds when the car requests you to.

I understand the levels, but you are trying a semantic trick "within seconds" to discredit the value of level 3.

It is obvious that one need several/many seconds to hand over driving task safely if the driver is not intended to pay attention.
The car manufacturer will of course never deploy a system where the system, in the transition phase, let the car crash because of a principle. Such a system is worthless (here we might agree?) and will put huge liabilities on the car manufacturer.

The only way to interpret L3 is that the system safely stop the car if the driver doesn't respond within a reasonable interval. That interval is certainly closer to 10 seconds than "within seconds". The stopping procedure will probably then take another 10-20 seconds.

 

[EVNow]

Most of the serious contenders actually building self-driving systems think the levels are stupid and even a bit evil because of how they mislead the public about a technology that is just barely deployed, and certainly wasn't when the levels were made up.

Don't say levels are stupid - @diplomat33 will be offended. I believe the new term is idiotic

The idea of levels suggested that it was just a progression of levels. I convinced SAE to put words in the document saying that they were not, and but even so people ignore them.

Very interesting. When the term is "level" followed by a number - obviously people are going to think it is progressive and 3 is better than 2 etc., irrespective of the fine print. You might say just like people get confused with the term "FSD"

Most, except for Elon Musk, think that ADAS and self-driving are two different things, and you don't just make self-driving by improving ADAS until it's super good. Former Tesla Autopilot lead Sterling Anderson described that approach as trying to build a ladder to the moon.

Actually Cruise thinks its a very valid approach.

04:48 : so there are a couple different ways that you can approach that problem you can start with a really low cost system like an L2 system and try to work your way up from a performance point of view or you can take the approach that we've taken​

Some interesting things the Cruise CEO said just a few months back that is opposite of what Waymo fans here insist is the "only way" to make FSD work. And he acknowledges that using multiple sensors, HD Maps and Geofencing are all temporary measures and the ultimate goal is not achievable with these.

04:26 : okay so let's take a step back and just expand a bit more on our overall approach to autonomy the goal obviously in the medium to longer term for everybody working on this is to get to low-cost generalized autonomy we want fully autonomous driving we want it to work everywhere and we want it to be really inexpensive

04:48 : so there are a couple different ways that you can approach that problem you can start with a really low cost system like an L2 system and try to work your way up from a performance point of view or you can take the approach that we've taken

05:31 in order to have a shot at solving that we then relaxed a bunch of other constraints so we said we're going to throw as much compute as we possibly can at the problem we're going to throw as many sensors as we need to at the problem we're going to use high definition maps to help get there and we're going to put in place a geofence if we need to sort of optimize around that ...


View attachment 771817

Wow ... so what do we have here ? No - he is not "gloating" about multiple sensors and HD Maps ... he says they are "relaxing constraints" (a.k.a compromises / crutches they are using). He clearly understands and says they took all these short term measures in order to get to 1000x better at driving faster.
- Lots of sensors
- HD Maps
- Geofencing

But some want the world to believe that crutches are actually great and indeed the only way to get to FSD - looks like the classic "make a virtue of necessity".

View attachment 771824

 

[EVNow]

I understand the levels, but you are trying a semantic trick "within seconds" to discredit the value of level 3.

Nothing wrong with L3 per say ... its the ODD that creates problems. Ofcourse the UN standard that is adopted in countries like Germany that Mercedes L3 will be based on, which stipulates "5 minutes" of testing, makes a mockery of the whole system.

The car manufacturer will of course never deploy a system where the system, in the transition phase, let the car crash because of a principle. Such a system is worthless (here we might agree?) and will put huge liabilities on the car manufacturer.

Wrong. See Honda Legend that I've commented on before.

The problem is - a manufacturer can put out a L3 system in a very expensive car that sells only a few dozen and claim they are "ahead" of a competitor. Clueless media will tout it and clueless people will start threads here asking why is Tesla behind

The only way to interpret L3 is that the system safely stop the car if the driver doesn't respond within a reasonable interval. That interval is certainly closer to 10 seconds than "within seconds". The stopping procedure will probably then take another 10-20 seconds.

You can't interpret however you want when standards don't specify that - including the UN regulation 157 that Merc is basing their L3 on (which Germany has adopted). You can read the standard - there is nothing there about how to safely stop the car after giving the driver 10 seconds (IIRC).

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

unece.org

Wiki - Consumer AV - Status Tracking Thread

 

[Daniel in SD]

The idea of levels suggested that it was just a progression of levels. I convinced SAE to put words in the document saying that they were not, and but even so people ignore them.

Good article. It’s too bad you couldn’t have convinced them to not use the term levels at all.
Is there another critique of SAE J3016 or is it all about the implied progression through “levels”?

 

[Daniel in SD]

@Daniel in SD has made it his life's "goal" to prevent others from being able to use FSD Beta. He has written about it explicitly. So, he wants to prove that Tesla is "illegally testing autonomous vehicles" and to get CA DMV to stop them doing it by making FSD Beta illegal.

Not true. I just like to argue on the internet and I think FSD Beta is an autonomous vehicle prototype. I actually think they should do away with disengagement reporting requirements entirely because they tell us nothing about how safe testing these systems on public roads is. I have never said that autonomous vehicle testing should be illegal, I’m a huge advocate of it. I just think FSD Beta should be subject to the same oversight that other autonomous vehicle testing is.

 

[Knightshade]

The only way to interpret L3 is that the system safely stop the car if the driver doesn't respond within a reasonable interval.

If the car can always stop safely without a human as the backup system then it's L4, not L3.

That's kind of the defining difference in the systems- the ability to always perform the backup "minimal risk" task without requiring a human to do so.

Which is another reason I find L3 so goofy....because it seems like "can do the entire DDT" is a much harder ask than "can safely pull over and park"

Not true. I just like to argue on the internet and I think FSD Beta is an autonomous vehicle prototype.

As someone who also enjoys arguing on the internet, I admire the tenacity with which you insist on being wrong about a point even Tesla own internal documentation tells you you're wrong about.

 

[bradtem]

Good article. It’s too bad you couldn’t have convinced them to not use the term levels at all.
Is there another critique of SAE J3016 or is it all about the implied progression through “levels”?

There are various critiques of J3016, but the core critique is not over their particular definitions per se, but that they were trying to solve the wrong problem. NHTSA invented a taxonomy of self-driving car tech with the primary characteristic being the role of the human being in operation of the system. This is like defining levels of motorwagen based on the role of the horse, as I satirize Uncovered: NHTSA Levels of 1900 (Satire) here. It's the wrong thing whether you do it well or badly.

The mistake is in thinking you can define a taxonomy for a technology still being developed that would be meaningful once it is deployed. Or even while it is being developed. The most useful thing to come out of those documents was instead the reminder that the vehicles don't work everywhere and all the time, that they have, in the term they used for it, an O.D.D. NHTSA's levels didn't even have a fifth level, and level 5 is not real, it's just a science fiction goal to remind you that the real robocars don't where on every street the way a human sort of does.

For most developers there has always been only one goal, which we call a self-driving car or robocar or whatever. A car that, on a reasonable and useful subset of streets and situations, can drive with no human involved. Everything else is ... something else.

 

[Knightshade]

and level 5 is not real,

Sorry, made me think of this...

https://twitter.com/x/status/1502369130887979016