Hacker News new | past | comments | ask | show | jobs | submit login
As of April 7, 2020, CA DMV has issued two Driverless Vehicle Testing Permits (ca.gov)
108 points by pengaru on April 13, 2020 | hide | past | favorite | 60 comments



Interesting. There are permits for "testing with a driver" (about 20 of those), and "testing without a driver" (only 2), "deployment for public use" (none so far), and "autonomous delivery trucks" (none listed).

The CA DMV's requirements for testing with a driver read a lot like those for a learner's permit - must have licensed driver in vehicle, no commercial use, no large vehicles. Testing without a driver has more requirements, although it's mostly self-certification. DMV doesn't require a road test.

Here's the form for applying for testing without a driver.[1] It's all pretty straightforward. If you actually have technology to do the job reliably, it's not going to be a problem. Uber used to complain that California was too restrictive for them to test here, until they ran down a pedestrian in Arizona. DMV seems to be doing a good job of sorting out the working technology from the fake-it-til-you-make-it crowd.

[1] https://www.dmv.ca.gov/portal/wcm/connect/63d33316-895a-474c...


Are there really only 20 permits for self-driving cars with drivers?

I swear there are more than 20 self-driving cars driving around the streets of Silicon Valley.


20 permitted companies


In case anyone is interested in a permit, they run a cool $3600, renewed every 2 years.

https://www.dmv.ca.gov/portal/dmv/?1dmy&urile=wcm:path:/dmv_...

you also need to own more than 25 registered motor vehicles and a five million dollar surety bond.

interesting bits...

> a manufacturer shall report within 10 days after the collision, any collision originating from the operation of the autonomous vehicle that resulted in property damage, bodily injury or death.

10 days? why not 10 hours? these manufacturers ideally know at all times where each vehicle is and what its doing, and a police report takes about 20 minutes.


10 days is the California standard time limit for reporting accidents: https://www.dmv.ca.gov/portal/dmv/detail/forms/sr/sr1

This imposes the same rules on machines / manufacturers as humans. Seems right for this phase of development. If this whole driverless cars thing works out hopefully the machines can do better than humans. They can sit on hold for longer at least.


Bonds (IMO) are similar to fast food franchise fees. They do a decent job of weeding out the majority of those that cut corners. 25 doesn’t seem all that high....Would be interesting to see what the # of registered vehicles is for each Fortune 500 (would be curious to see if dmv sold that info).


you also need to own more than 25 registered motor vehicles and a five million dollar surety bond.

The OL 311 application form and 13 CCR § 227.04(c) say you have to provide one of insurance, self-insurance, or a surety bond. The requirement to have 25 vehicles applies only to self-insurance, as for any other self-insurer.


because 10 days is the same amount of time you have to report a personal crash as well


PR team has to work around the clock on how to spin this. People really blow up anything related to self driving, even when others are at fault most of the time.


For those curious:

Waymo is owned by Google (well, Alphabet), and is looking to compete with Uber/Lyft.

Nuro is backed by Softbank, and is looking to start with food/grocery deliveries.


Waymo has more competition with Cruise, Aurora, and Argo than what Uber/Lyft is trying.

I've heard that Uber ATG is barely staying afloat and Lyft is just pumping money just as a desperate attempt (their platform is much more powerful, which is what they're betting on).


How is lyft's platform more powerful? Their apps both seem identical.


Lyft is opening up their app for key players: They opened it up for Waymo and are working with others (i forgot their names) to also put them on their platform.


Well, what I mean is the ultimate goal of Waymo is on-demand rides from self driving cars, much like what Uber/Lyft do today with humans (rather than, for example, selling the cars).


I interviewed with one of those companies and it just seems bizarre that people are investing so many millions of dollars into building a taxi company. How big is the taxi industry when all is said and done?


They aren't replacing taxis, they are replacing cars.

Unless you want to own a specific speciality car (Ferrari, Jeep, etc), Waymo eliminates the need for you to buy one. It's Uber but without the overhead of paying a person, which means price could normalize so close to the actual operating expenses of the vehicle that it just doesn't make financial sense to use anything else.



That study doesn't seem terribly relevant. It concludes that owning a vehicle is important to people right now. It doesn't seem unreasonable to assume that in the future, cheap and ubiquitous ride sharing options could pretty quickly drive those percentages down.


Well, if I'm reading this right (https://investor.uber.com/news-events/news/press-release-det...), Uber's share of it alone brought in $65 billion in gross bookings last year. So, not exactly what I'd call small.

What they need is to figure out how to massively decrease operating margins, and for many years now the story has been how it'll be done - at least partly - with autonomous vehicles. I'm skeptical, but we'll see. It would make a lot more sense to me if the robotaxis were used only on certain roads outfitted with multiple types of redundant roadside sensors to assist the self-driving cars.


They're not taxi companies, they're tech companies. At least, until people listen to the little kid asking where's the clothes and everybody realizes the deceit (as happened to WeWork).


They’re down 40% from IPO so there appear to be quite a few people spotting nudity.


They hope that if they perfect autonomous vehicle technology that will allow them to both expand the taxi market, and license the technology to every other vehicle manufacturer.


Can you imagine a more captive audience for ads than those in a cab? Imagine Alphabet introducing free, ad supported taxi rides.


So you'd be forbidden to look at your phone or wear headphones during the ride?


I am sure if you are offered the choice between gamified ads and paying for the ride more than a few would rather waste some time interacting with an ad. Especially because nothing stops them from charging exorbitant cash rates. The app would disclose it , of course: this ride costs you a hundred dollars but if you want you can play Awesome Coke Hunting instead...


Kind of surprised Tesla isn’t further along. Maybe FSD is further than we thought.


Owner here of a TM3 here.

To me the issue is that Tesla seems to be shooting for the moon, trying to do it all in one shot, rather than its on rails method that we have seen from other manufacturers. By on rails I mean only on very accurately mapped routes with no deviation.

I would not trust my car with my eyes closed, I do however trust it with supervision which still is a major relief from the stress of driving.

One issue that seems being dodged is, my current software and hardware allows the cars too represent stop signs, traffic lights, cones, and oddly trashcans. Guess what is missing, speed limit signs. The suspicion is that MobileEye holds the patent and either Tesla won't pay for it or MobileEye won't license it. So how you can have a self driving car that cannot recognize speed limit signs is not evident to me; road data is not complete.


That can be part of the maps data, right? EG Google maps shows the speed limit when navigating.


What about variable speed limits, construction, etc?


Tesla don't need one of these because they're starting out "with driver" for FSD. Like they do for AutoPilot, the driver is expected to take control if the vehicle is doing something unexpected.


Elon Musk is the Elizabeth Holmes of driverless cars.


Musk tweeted today "Functionality still looking good for this year. Regulatory approval is the big unknown"

With regards to the "first robo taxi deployment"

[1] https://twitter.com/elonmusk/status/1249210220200550405?s=21


It's worth noting that Musk has previously claimed that Tesla would have fully autonomous vehicles by 2017, by mid 2019, by the end of 2019, by mid 2020, and now apparently by the end of 2020, pending regulatory approval.

I'm not convinced.


Also worth noting he claimed they will do first fly of Falcon 9 in 2007, but happened in 2010. Same with Model3 production rates etc. Also worth noting that others just don't provide any time frames in fear of failing those. Musk is just more open with plans and often too optimistic, but they still have most advanced self driving capabilities that are actually mass selling now.


> still have most advanced self driving capabilities that are actually mass selling now.

But they don't have autonomy. They have driver assist. And this bad marketing gets people killed.


I didn't say they have. But autonomy doesn't mean immunity from fatalities as well. Their cars as it is already lower overall fatality rate.


But they don't have autonomy. Maybe they lower fatality rate but they are not anywhere near Waymo (or Cruise for that matter) if you look at the metrics.


It's weird that I have to point out second time I _didn't say they have autonomy_. And what's wrong with lowering fatality rate whatever your competitors are doing? Also, I'm anxious to see Waymo released, but current reality is what it is, Waymo is still in testing.


> And what's wrong with lowering fatality rate whatever your competitors are doing?

It's not actually clear that Tesla lowers the fatality rate. Tesla's driver assist features only work on highways, and highway miles are some of the safest. But the data you get compares assisted teslas on highways to average drivers on all roads. And it's not clear that Tesla is categorically better than competitor's driver assist technology (https://qz.com/1414132/teslas-first-accident-report-claims-i...).

So nothing you've claimed is strongly supported by evidence, and given that these are safety claims that you're parroting from tesla, I'm going to reiterate: Tesla's claims of improved safety and marketing are dangerous since they aren't true.

> It's weird that I have to point out second time I _didn't say they have autonomy_.

Right, but Tesla does, they call their driver assist "Autopilot".


Autopilot != autonomy in cars or in planes.


I'm not sure what you're getting at. In planes (and especially in planes in pop culture), autopilot does allow flight without keeping your hands on the wheel and eyes on the road, so to speak.


If you are noting deadlines, how about ones else beat? Model Y was 6 months early.


That's cool. My point was that Musk is not afraid to say optimistic time frames based on current expectations, sometimes they work, sometimes they don't. I prefer it much more than the usual way corporations work - play safe and don't promise anything.


> Musk is just more open with plans and often too optimistic,

Or call it what it is, lying. He habitually lies about timelines.

> still have most advanced self driving capabilities that are actually mass selling now.

Easy to be the most advanced if you're the only one "selling".


I wouldn't go so far as to call it lying. Unless you believe anyone who makes an inaccurate forward looking statement is lying, in which case I would say the majority of people lie the majority of the time when making forward looking statements.


It's a lie because he knows he can't reasonably achieve anything he's promising on any of the stated timelines. It's Theranos but with cars.


It's possible. Do you have any specific examples?

My experience with forward looking statements has been that people interpret something with many caveats as promises with no caveats.


It's not lying if he believes what he says. Do you have proofs that he deliberately does it?

> Easy to be the most advanced if you're the only one "selling". I don't understand what you mean with quotes.


The constant pattern of over promising and at most under delivering indicates shenanigans.

Telsa's FSD package is a scam that will never be delivered. It's not a sale because there's no service delivered, it's a lie meant to trick people and investors out of their money.


They also overdeliver sometimes.


This is true, no one expected a fart app for a car.


He also fully and openly admits his timing is often wrong. Specifically for this exact target actually, during the big investor meeting for it, he said he is sure they'll be able to do this but the time estimate could be off.


Because lying to investors can get you in trouble.


It's also worth noting that rockets have been done before.


Not with landing of the booster, which is the whole point of Falcon 9.


I am pretty sure Musk runs on Mars time. One year on earth are 1,9 years on Mars ;)


Another way to look at it is Musk has been right about engineering/manufacturing goals but has consistently missed on AI goals.


They released the Model Y ahead of schedule.


*for certain idiosyncratic definitions of “fully autonomous”


Can @dang or someone update the title with "California DMV" to be a bit more specific here?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: