Uber won’t renew its permit to test self-driving cars in California

Photo by Vjeran Pavic / The Verge

Uber will not renew its permit to test autonomous vehicles on public roads in California. It’s further evidence that the company is pulling back from its aggressive plans to launch a self-driving service in the wake of a fatal crash in Arizona, which resulted in the death of a pedestrian.

In a letter, the California DMV confirmed that Uber’s authority to test self-driving cars in the state will end March 31st following the decision not to renew its license. A spokesperson for Uber confirmed that it would not seek to continue testing in California in light of the crash in Arizona. Federal investigators are currently probing the cause of the crash.

“We proactively suspended our self-driving operations, including in California, immediately following the Tempe incident,” the spokesperson said. “Given this, we decided to not reapply for a California DMV permit with the understanding that our self-driving vehicles would not operate on public roads in the immediate future.”

Arizona governor Doug Ducey suspended Uber “indefinitely” from testing in the state, after describing the dash cam footage of the crash released by the Tempe Police Department as “disturbing.” Uber has also grounded its self-driving vehicles in Pittsburgh and Toronto.

The story of Uber’s permit to test autonomous vehicles in California is a saga unto itself. The company infamously began testing its self-driving cars in San Francisco in December 2016 in defiance of the DMV’s order for the company to obtain the permit. Uber refused, and the DMV revoked the licenses of its self-driving fleet. That prompted the company to move its testing operation to Tempe, where almost a year later, a pedestrian was killed while crossing in front of a self-driving Uber vehicle.

According to emails obtained by The Verge, Uber knew for months that its self-driving cars violated the state’s requirement that autonomous vehicles be licensed before operating on public roads. Yet, it still deployed its fleet of autonomous vehicles anyway, arguing that because they required a human driver to monitor the vehicle at all times, they did not legally meet the definition of an autonomous vehicle under the state’s law. The company did eventually conceded, obtaining a permit in March 2017.

Uber clearly wanted to avoid a political fight with the DMV that a permit renewal likely would have entailed. The agency is gearing up to begin issuing permits in April to companies wishing to test fully driverless vehicles on public roads without human safety drivers.

Asked whether that program was still going forward in the wake of the Uber crash, a DMV spokesperson said, “The DMV is allowed to begin issuing driverless testing and/or deployment permits on April 2, but that doesn’t mean a manufacturer will meet the requirements or if we will approve them.” In other words, Uber’s chances of receiving a permit were probably slim to none.

The fate of Uber’s self-driving program very much hangs in the balance. According to a report in The New York Times, the goal of offering driverless ride-hailing services to the general public by the end of the year was quickly falling apart, and Uber’s self-driving cars had a record of failing to operate correctly under a number of standard road conditions. Uber CEO Dara Khosrowshahi mulled whether to pull the plug before eventually deciding to continue the program.


Uber clearly under-estimated the difficulty of building reliable SDCs
It’s easy to forget Waymo’s program started almost a decade ago and have accumulated huge amount of heuristics, that’s not something you can just make up with more training data.

Same can be said for Tesla, which is arguably in an even worse position, i.e. using same platform as Uber but without lidar…

"LiDAR costs money is a crutch. I want to drive over people in the dark by only relying on cameras"

- Elon Musk

Which is totally possible! It’s just a longer path to Level 4+ because they tech can’t compete with LIDAR’s accuracy yet.

I think "yet" might be embellishing a bit.

There are "cameras" that can see well outside "visible" spectrum. these different spectrum could improve night vision or seeing in rain/fog. Remember self-driving tech is super new, and people are super bad drivers… the bar is pretty low.

But those aren’t being used. Elon’s shtick is black and white cameras with sonar/radar. The reason he is going this route is because of the cheap costs involved and the ease of converting anything to autonomy if and when they figure it out. This arguably is not working.

Not really. See this 2010 paper on using overhead imaging and a stereo pair to get location accuracy of about 10m, comparable to standard GPS on the vehicle.
Also this 2015 paper which uses a 3D LIDAR map and a monocular camera for cm level accuracy.
Both are relatively quite old and were done using far fewer vehicles than Tesla has data from.

Keep in mind that I’m thinking Level 5 LIDAR based vehicles in 5-10 years, with camera/RADAR vehicles another 5-10 years behind that.

Will this be a lesson for Uber in playing fast and loose? Probably not, but it will be an interesting case study in business textbooks for years to come.

I really hope it is. If it isn’t we should implement new regulations to force them to not play fast and loose.

Looks like Alphabet is in the driving seat now. Slow and steady is the way to go in this industry and Waymo might be Alphabet next big hit that they have been waiting for.

There’s still GM and Ford, along with plenty of others. Waymo’s more public (partially because of the court case) but there’s definitely a race going on out there.

Not to many can compete with Waymo though. Probably just GM and Ford at this point.

I think there are too many competitors in the market for Waymo to be a dominant player in driver-less car market. Most of people can agree Waymo has the most advanced self driving car technology, but other car makers know how it will effect their business model and already are working on their own driver-less car system, not to mention numerous self driving car start ups that are cooperating with car makers.
But who knows? Waymo’s technology could be WAY better and every car makers would implement its system .

This unfortunate fatal accident might of happened to Uber’s self driving test program. In the end, this accident simply put all self driving car test company on hot water. If another accident happens, hopefully not resulting in fatality, you pretty much save to say all self driving car programs will be place on hold indefinitely and states will think twice before giving out any permits again.

Yup, Nvidia is already "temporarily suspending the testing of our self-driving cars on public roads to learn from the Uber Incident,"

This will affect everyone in the self driving space.

IMO if any company feels confident enough to put their cars on the road for autonomous testing, the only affect this will have is public acceptance, otherwise, they shouldn’t have been publicly testing in the first place.

Waymo has Intel inside.

There is too many problems today with self driving cars. Computers are not powerful enough. Our brains do a lot while we are driving, mostly we do not pay attention what is happening. We see a red light a block down the street, at the moment it would have low priority, kids playing ball on the side of the road, we carefully pass as we noticed brake lights on a parked car come on, we are ready for it too pull out just in front of us, but now the red light is still red an we start slowing down as it turns green, we still don’t rush knowing there are red-light runners. As we cross the street, we remember that large pothole that we have to drive around. The rain has now started and the poorly marked lanes have disappeared… But we will never know for sure what the car next to us is going to do.

Self Driving Cars is nothing new. On I-8 as I recall over 50 years ago there was talk on making part of it a test zone with a buried cable in the road that the car would follow, never materialized.

We have some cool features today we did not have in the past to help us stop and park as well as tell us how to get somewhere, maybe.

The biggest concern about auto safety is the nut that holds the steering wheel.

IMO… Not enough credit is going to the State of California’s Department of Transportation and their tough stance on safety requirements and reporting.
All we heard from Uber for 2 years was that Cali was too strict and required too much information and that it would be too hard to comply with their regulations ‘so we’re going to find another state that doesn’t care as much and test our world-changing software there and you’ll miss out on our investement & innovation. Wah.’
All they ask is for every self-driving company to at least meet a minimum requirement for testing and provide adequate testing results. I’m not saying this accident couldn’t of occured in California or that the requirements would of rendered it preventable, but its more to the fact that when everyone has a vested interest in safety, (and safety as the leading priority) that it shows and it bears out in ways that make the company not want to cut corners or rush to market.
When you let companies decide what should be prioritized, it’ll almost always be profit.
When you ask them to police themselves, we all get screwed…
…Just ask Facebook.

(I don’t normally do this, but i copied & pasted my comment from yesterday’s Uber story. It fits better here)

Didn’t Google just get a big stake in Uber from their lawsuit? Wouldn’t it make sense to let Uber’s value continue to crash and buy the whole company? That way Waymo self driving vans would have access to Uber customers and all their customer and driver information. It’d help them expand nationwide a lot faster.

Uber is worth north of seventy billion at this point. Google got 280’ish million in stock.
edit: Was worth. Probably a bit less now, considering the fatality.

Yeah nevermind. I thought they had a bigger stake in Uber than that. I still wouldn’t be surprised to see them at least consider buying Uber to get their self driving vans nationwide as fast as possible.

What a shame, it definitely will slow overall progress in this field. Autonomous cars are already way way safer than humans. There are around 1000 fatalities on the roads of Arizona caused by drivers and pedestrians, yet no one is banning cars. Each year there are thousands of deaths and injuries on the roads just because we let people drive, instead of computers.

And people really fear computers driving them around. Hell, even on this site, I’ve seen people say self driving tech is overstating how accurate it is and blah blah blah. People won’t accept self driving tech right away, no matter how safe it is. That’s just a fact of life. Wait til they put it out in the wild. Myself and others would be the first to ride in them. And everyone else would gradually grow to accept it. Or at least the majority would.

View All Comments
Back to top ↑