Volvo has promised to sell uber cars. That's pretty much the gist of it.
Uber got a fleet quote, attached the words self-driving car to it, then issued a press release.
Volvo's part in the self-driving car is that their cars meet the minimum specs for uber's program, which are probably more stringent than "must have 4 wheels" and yet still falls far short of "must have a lidar system accessible through an api".
Crackpot realism is one of the downsides of the division of labor. It emerges reliably whenever two conditions are in effect. The first condition is that the task of choosing goals for an activity is assigned to one group of people and the task of finding means to achieve those goals is left to a different group of people. The second condition is that the first group needs to be enough higher in social status than the second group that members of the first group need pay no attention to the concerns of the second group.
Consider, as an example, the plight of a team of engineers tasked with designing a flying car. People have been trying to do this for more than a century now, and the results are in: it’s a really dumb idea. It so happens that a great many of the engineering features that make a good car make a bad aircraft, and vice versa; for instance, an auto engine needs to be optimized for torque rather than speed, while an aircraft engine needs to be optimized for speed rather than torque. Thus every flying car ever built—and there have been plenty of them—performed just as poorly as a car as it did as a plane, and cost so much that for the same price you could buy a good car, a good airplane, and enough fuel to keep both of them running for a good long time.
Engineers know this. Still, if you’re an engineer and you’ve been hired by some clueless tech-industry godzillionaire who wants a flying car, you probably don’t have the option of telling your employer the truth about his pet project—that is, that no matter how much of his money he plows into the project, he’s going to get a clunker of a vehicle that won’t be any good at either of its two incompatible roles—because he’ll simply fire you and hire someone who will tell him what he wants to hear. Nor do you have the option of sitting him down and getting him to face what’s behind his own unexamined desires and expectations, so that he might notice that his fixation on having a flying car is an emotionally charged hangover from age eight, when he daydreamed about having one to help him cope with the miserable, bully-ridden public school system in which he was trapped for so many wretched years. So you devote your working hours to finding the most rational, scientific, and utilitarian means to accomplish a pointless, useless, and self-defeating end. That’s crackpot realism.
Well call it a personal air vehicle or affordable helicopter. It isn't going to drive on the roads so it's not a flying car in that sense
But otherwise, there is some validity in the criticisms. In the tech world, the management usually a lot behind the curve on technologies compared to the more hands on people.
I'm won't be surprised to see self driving ubers within a few years, however I'll be amazed if I see driverless Ubers.
This is why I think at launch these will have non-drivers at the drivers seat smiling and ready to take over if the AI refuses. These drivers will have to drive less and less, and will eventually be remote operators. How they will get any economy in that I don't know - I'm cynically thinking they will be doing telemarketing while idling behind the wheel.
> however I'll be amazed if I see driverless Ubers
I won't. Uber knows exactly where you want to start and end, time of day, the route to take, and the historical performance of that route (eg speeds, accident rates, variability). Consequently even if the driverless cars are only good for 1% of routes, they would be able to use them. Soon 1% becomes 2% etc.
That might be another option, but then you risk the car getting stuck on its way to the customer. Also, I want to be able to Uber home from a bar drunk like an idiot, or take an uber because I can't drive.
I definitely see a place for share pool cars that deliver themselves to you when you need them and return autonomously when you are done, but maybe not Uber?
Would this mean that the "driver" seat is an open seat (intentionally leaving the accessible the controls)? This seems interesting, but legally would this be problematic? Who is at fault when the malfunctions and the user tries to help yet the accidental is most likely unavoidable given the disconnect between being the driver and being the stand-by driver? I would think that your reflexes and lack of actually controlling the car would be harmful to reaction time.
You answered it yourself: and will eventually be remote operators.
If the cars need a human 50% of the time, and you move the operators to a call center, you can get away with fewer drivers (probably quite a bit more than half the number because of queueing theory). You also need to pay drivers less if they can work from home (they won’t lose working hours driving empty cars) You will also, likely, use less fuel, especially if you make your cars smaller.
Has anyone considered the possibility that people won't want to trust their lives to self driving cars? We've had self checkouts at the supermarket for ages now and they still give problems with unusual situations. When people think of the infamous "please put your item in the bagging area" might they not get nervous about trusting their life to soemthing else autonomous?
The meaning of self in self driving and self checkout is different. They are more or less the exact opposite. In the first case the machine is doing something itself, in the other case you do something yourself.
Living in an area with self-driving ubers running about; I think there will be a checkbox on the app, 'Self Driving?' If you click Yes, could get a self driving car (and it'll be faster/cheaper/safer), click No and it'll be the same experience as previous (though probably with a bit more lag as self driving will crowd out a bit of the existing car fleet).
For at least a number of years it'll be more supplementary than anything for areas with low Uber driver counts.
As a system an elevator is working in a infinitely more controlled environment when compared to a driverless car please consider these problem areas:
-Localisation and perception in all matter of conditions e.g. terrain or wheather changes, interaction with traffic, pedestrian behaviour, unexpected or unclassified agents (road works, car stopped in the middle lane of a highway)
- Vehicle dynamics sensing, vehicle diagnostics. Any bog standard car will have in excess of 30+ control units (from micro controllers to full on embedded computers). How do you keep tabs on all of them, which one's can you trust, if my LIDAR says I'm good but my RADAR false triggers on a low beam tunnel which one is accurate? Should I slam the brakes or should I drive through what could be a front collision?
- Actuation, how smooth is smooth? I can be within the speed limit and stil scare the crap out of you driving above an "adequate" speed for the environment, you can brake with an ABS pump far better than a human can and avoid a collision at all cost, but the G forces involved would probably snap the neck of everyone in the car, what would be the acceptable outcome in that instance?
There is a lot of complexity in all of this and realistically the only way you can trust the current levels of driverless is by limiting/controlling the environments they are allowed to be enabled in.
Most of those kinds of bagging systems have gotten much better about the whole "detection of bagging."
However, most of the time, the issue with self-checkout is the HUMAN, not the machine. Several extremely busy grocery stores near me ripped out self-checkout systems because it affected throughput too much. They went back to cashiers because the cashiers are A) faster and B) capable of judging situations for throughput ("Um, grandma, really, we don't need you to take 5 minutes to find two pennies for that $20.02 grocery bill. Have a nice day. NEXT!")
Amazon, unusually, has this one right. A self-checkout system CANNOT rely on the customer doing anything right.
Having a few self-checkout stations in a grocery store is generally OK if they're effectively for people who have dashed into the store to pick up one of two barcoded items. They're horrible when the store has used them to replace most of their cashiers because they're slow to use for a lot of items, are very slow to use for produce, etc.
I still find detection of bagging is often a pain with self-checkout. I expect things are set up to err on the side of falsely detecting potentially unpaid items.
Perhaps, at some point, vision systems will make it easier to not require cashiers but it doesn't work all that well today for large supermarkets.
If you are buying more than a small handful of things, they continually beep and complain about the bagging area as things shift around or have to get moved to make room. And even worse if you are using your own bags.
Many people will not trust them. Within a few years though there will be hard data showing that the human drivers are significantly less safe and reliable and at least some places will outlaw human drivers.
> Has anyone considered the possibility that people won't want to trust their lives to self driving cars?
Has anyone considered that the average driver sucks? And that 50% of the drivers are worse than that?
Self-driving cars have a very low bar to clear (better than average human), and they will probably clear that fairly soon given just how bad normal drivers are.
Once self-driving cars are deployed, it's simply a matter of crunching the data on more and more unusual situations.
Well, good luck
First, the insurer needs to have a lawful decision on who is guilty - it is not their job to prove anything. And it's still you vs huge corporation.
Looking at what these companies (Tesla for example) do to people criticizing their products - full spectrum, from lie accusations through personal attacks to law threats, i'd say even if you have a recording the situation is still bad.
So, we're going to get a confirmation what you just said - their lawyers will do everything to prove you suck as a driver.
The XC90 comes with a pretty good LKAS and ACC system, having recently test driven every car in that class, I can say it's the best one for under $60k (Highlander is pretty good too). But there's no way that Uber is going to be using the stock system for self-driving.
Wonder how well they will do with flooded roads. Will they turn around and not drown.
My significant other is a lyft driver, the other night she had a drunk rider and he became unresponsive. And fell on the flood of the van. She had to stop and check him. Get back back into his seat and at his destination go knock on the door for people to come get him. Would the driverless car do all that?
How well are the on snow covered roads as well? Anyone road in one is heavy snow like conditions?
Hidden behind a paywall, is it possible to get a mirror? I'm curious to see if it will be self-drive or assisted driving. Volvo thinks they can create the tech and mass produce 24,000 in the next 2 years, so how's their progress so far?
>“The base vehicles incorporate all necessary safety, redundancy and core autonomous driving technologies that are required for Uber to add its own self-driving technology,” Mr. Samuelsson said. “Uber will adapt the software to make it ride-hailing.”
In other words, it appears the vehicles will have some level of autonomous/assistive driving systems but it will be an exercise for Uber to actually make them fully autonomous. (Which seems highly unlikely in that timeframe.)
By William Boston
Updated Nov. 20, 2017 12:22 p.m. ET
71 COMMENTS
Volvo Cars said it has agreed to supply Uber Technologies Inc. with a fleet of 24,000 self-driving taxis beginning in 2019—one of the first and biggest commercial orders for such vehicles.
The deal between Volvo, owned by China’s Zhejiang Geely Holding Group Co., and Uber was disclosed Monday as a framework agreement without financial terms. Such an order, though, would account for about 4.5% of Volvo’s current total sales, based on 2016 figures, and is estimated to be worth just over $1 billion, a person familiar with the situation said.
The automobile industry and many big tech firms have promised broad deployment of robot cars in the near future, but all of the technology, regulation and legal framework needed for the practical use of such vehicles aren’t yet in place. Still, industry executives have predicted with more and more confidence that companies could have such cars on the road in a matter of years, not decades.
Other car makers have inked smaller, less specific deals to develop and produce autonomous cars, but Monday’s agreement represents one of the most concrete deals between two big players in the field for the production of a large number of real cars. The promised delivery date--as little as two years away--is also one of the first hard, deadlines that a significant automaker has set for rolling out a working model.
Volvo has had a long-running partnership with Uber to develop self-driving cars, though it hit a setback earlier this year. Pilot projects around the U.S. were temporarily halted after an early version of a Volvo-Uber-developed vehicle flipped on its side after an accident in Tempe, Arizona. Police said the incident was caused by a human driver of another vehicle, not the Uber self-driving car.
“We believe this is a new segment, a new business,” said Volvo Cars Chief Executive Hakan Samuelsson. Jeff Miller, Uber’s head of auto alliances, said the agreement “puts us on a path towards mass produced self-driving vehicles at scale.”
Under the agreement, Volvo will supply Uber with what Mr. Samuelsson called a “base car,” based on Volvo’s popular XC90, a luxury sport-utility vehicle that seats up to seven passengers. Delivery of the vehicles is set to begin in 2019, with Uber calling up vehicles each month through 2021.
“The base vehicles incorporate all necessary safety, redundancy and core autonomous driving technologies that are required for Uber to add its own self-driving technology,” Mr. Samuelsson said. “Uber will adapt the software to make it ride-hailing.”
The first taxis from the deal will be built in Volvo’s factory in Sweden. But the company plans to produce the vehicles in the U.S. as well, where Volvo is building a new factory just outside Charleston, South Carolina.
The current version of the XC90, which is available for sale today, already includes advance autonomous features, such as systems to keep the car in its lane and maintain the proper distance to another vehicle traveling ahead. It also features collision avoidance that helps prevent low-speed fender benders in stop-and-go traffic.
Other car manufacturers are racing to develop their own models of self-driving vehicles, often in ventures with parts suppliers and big technology companies.
Why Your Next Car May Look Like a Living Room
With driverless cars moving closer to reality, car makers and designers are imagining a future where car interiors look more like a high-tech living room. Photo: Morgan Anderson/Yanfeng
Daimler AG, which makes Mercedes-Benz luxury cars, and auto supplier Bosch GmbH, for instance, said in April that they would jointly develop their own robot taxi. Daimler has said the vehicle would become available at the beginning of the next decade.
Daimler and Uber said in January they would join forces, with the German company agreeing to introduce self-driving cars compatible with Uber. Waymo, the self-driving tech unit of Google-parent Alphabet Inc., has agreed to take several hundred minivans from Fiat Chrysler Automobiles NV, part of its autonomous driving program.
Nutonomy Inc., a Cambridge, Mass. startup, has said it wants to put a fleet of robot taxis on the street by next year. It first began testing Renault and Mitsubishi vehicles, in Singapore. In May, it agreed to work with Peugeot maker Groupe PSA of France, to integrate its software into a Peugeot SUV.
Volvo and Uber agreed in 2016 to jointly develop more advanced self-driving car systems. The accident in Arizona was a blow to public perception of the pace of development of the program and such vehicles generally. But police said at the time that the Volvo-Uber car was not at fault. The accident occurred when the driver of another vehicle committed a moving violation, ramming the self-driving car and causing it to flip on its side.
There were no injuries, but Uber temporarily halted its three robot taxi tests in San Francisco, Tempe, and Pittsburgh. Those pilot programs have resumed.
—Tim Higgins in San Francisco contributed to this article.
But if I owned the WSJ (or any publication) and a subscriber shared an article here-and-there on a forum I personally wouldn't be mad. Perhaps doing so would even lead to more subscribers.
They're not. It appears that they've made a deal (without specific commitments) to sell Uber vehicles that Uber can turn into fully self-driving vehicles if it can. There would still appear to be some reputational risk but Volvo isn't making claims that the cars will be self-driving out of the factory.
autonomous cars will, with time, be safer then the human driven ones, so Volvo may also be playing here their traditional card of an early adopter of a new safety feature.
Don't expect these things to be even as good as Waymo's from day one...or first five years. And I don't even believe the reports that Waymo has gotten good enough for full autonomous driving in all environments.
Why Uber's pursuing a driverless model just baffles me. This quote is for 24,000 cars for around $1.4 billion, or around $58,000 per vehicle. Even if we assume 0 financing cost the car would need to earn over $58,000 in its lifetime to be more profitable for Uber to buy one of these than to continue to pay a driver. Current Uber drivers estimate their profit to be around $0.40 per mile[1] which means each driverless vehicle would need to drive about 150,000 miles just to break even.
Let's guess driving that much that takes about 2 years[2]--if you assume a conservative 1 technician per 50 vehicles[3] to maintain these you also have 2% of their salary * 2 years (say another $3K). Throw in another $1000 for tabs etc, $1500 for the labor for a couple hundred car washes[4], $1000+ for losses related to crashes[5], $500 for parts and maintenance[6] and we're easily at another $7K+ in costs that have to come out of Uber's existing share.
[1] https://therideshareguy.com/how-to-calculate-per-mile-earnin...
(This is for a 43MPG Prius, with ~$0.06/mile of depreciation. I exclude the depreciation for the calculation above, but need to add back in a similar amount for the difference in MPG.)
Nice references! Might be better to start with the cost paid by the customer to Uber rather than just the driver's cut of it?
"In San Francisco the fare is $2.20 plus $0.26 per minute plus $1.30 per mile. In New York City the fare is $3 plus $0.40 per minute plus $2.15 per mile."
That's not relevant because it doesn't change under a driverless system. Uber already gets "their" cut. The change is that Uber gains as costs what the driver would have paid to own, maintain, and operate the vehicle, and gains as an income what it would have paid the driver. The "driver profit per mile" nicely factors those things together.
I am not convinced self driving cars will ever be a thing we see in our lifetimes that just drive around the city unmanned. Sure, I can see a fully custom Uber navigational system built in with systems that make getting passengers and dropping them off much more predictable.
However, what should be relatively simple as that fails to address the human component.
1. Some people like talking to a driver, it's one of those small interactions that make getting from point A to point B an experience. Some people don't and that's fine too, not arguing against that.
2. People left alone will be much less careful, graffiti, spillage, throwing-up, drug deals, etc. Fully expect these to become little crime bubbles.
3. Not all cars will be fully utilized all day long, so what is a car sitting around unmanned? Will the insurance to cover vandalism and having the vehicle repaired raise higher than the value of paying someone to sit in the car?
1. The vast majority of people don't want getting to point A to point B to be an experience, they want it to happen as quickly and conveniently as possible for a good price. This isn't going to be any sort of real barrier to adoption especially when people have the option to pay $15 for a manned car with a driver or $3 for one that is self-driving.
2. Cameras and you sign an agreement on signup that you're liable for the behavior of any passengers on your trip. Hotels are also private spaces, they don't seem to have a problem with graffiti, rooms getting destroyed, etc. To the extent they do it's not an existential problem and is covered by the fact that they have your ID and Credit Card on file.
3. Many hundreds of thousands of people park their cars overnight in parking garages or on streets. Nobody pays people to sit inside inside of their cars overnight as far as I know.
Yeah. It's reasonable to be skeptical about the timeframe for fully-autonomous general purpose vehicles. IMO, the last 10 percent (or whatever) is tougher than a lot of people are assuming. Some of the assumptions about the economics of the vehicles are also overly optimistic.
That said, if people can get driven around for a reasonable price--either in their own vehicle or one they hail through an app--many people will forgo driving themselves.
The whole vandalism/making a mess thing is entirely possible to solve by holding the passenger accountable. I'd assume the cars would all have camera's inside them. The flow could be as simple as:
1) I use an app to request a ride
2) Car pulls up
3) Car does a facial recognition scan as I approach the car and know's it's me, unlocking the door. It know's it's me because I also had to do a facial recognition scan when I set up the app.
4) Camera's can monitor the cleanlyness of the car. If it detects a mess when I leave the vehicle then the car gets sent to the depot for a cleaning.
4.a) Car doesn't detect it's dirty, but the next passenger to get picked up notices it's a mess and can notify them via the app. This will prompt a new car to come for the passenger and the original vehicle gets sent to the depot for cleaning
5) The vehicle gets inspected and cleaned and whoever was responsible gets notifed of the cleaning fee and their credit card is billed automatically (which would be part of the agreement when using the service)
Good point. When I said "simply" it was more the fact that all this technology already exists. If you can fit facial ID scanners on a phone, surely you can on a self-driving car. And ff you know the clean state of the car, it's not very difficult technically to detect if anything was left in the car. This could even set a flag for some one to remotely check the camera to see if it's a false alarm. Again -- not very difficult on a technical standpoint. However, even bypassing this theoretical "auto-dirty-detector" you could soley rely on the next rider to mark the vehicle as unsanitary which would dispatch them a new ride, while the old one gets dispatched back to the depot for cleaning.
We've already seen how #1 goes with ATMs replacing bank tellers. Sure, some folks liked banking with a human, but they're largely replaced.
#2 you deal with by cameras in the car. You know who's in it, and you have a credit card to fine them for any clean-up required. Taxis/Ubers already have to deal with more puke than private cars. I wouldn't want to do a drug deal in a car that could automatically deliver me at a police precinct...
#3 is already the case with the millions of parked cars on city streets. Why would self-driving cars be more of a target than those existing unmanned cars?
> 1. Some people like talking to a driver, it's one of those small interactions that make getting from point A to point B an experience. Some people don't and that's fine too, not arguing against that.
Not many are going to care enough to pay a significant premium though.
> 2. People left alone will be much less careful, graffiti, spillage, throwing-up, drug deals, etc. Fully expect these to become little crime bubbles.
Expect self-driving cars to have surveillance systems and to require identification before you can hail one. If you're caught doing a drug deal, don't expect that company to ever let you use their service again.
> 3. Not all cars will be fully utilized all day long, so what is a car sitting around unmanned?
Park itself, probably.
> Will the insurance to cover vandalism and having the vehicle repaired raise higher than the value of paying someone to sit in the car?
Drug deals? You first have to be authorized - you have to give some payment information. So you are mostly easily identified. Then it will probably be under surveillance.
I believe Car2go had an issue with this. Identity theft is alive and well, and it's hard to know if the person who is in the car is the same who signed up for the account.
So they'll add some sort of biometric - face / iris / thumb.
People have better places to do drug deals than a car that can lock the doors and speed to the local police precinct (with photographic/video evidence sent in advance, even!) when it detects misbehavior.
Well, it did happen in our lifetime already. Waymo cars are driving beta customers around in Phoenix without anyone in the driving seat. There is still an employee in the back seat but within a few months they will not be in there at all.
So looks like your estimate was incorrect by many decades.
Uber got a fleet quote, attached the words self-driving car to it, then issued a press release.
Volvo's part in the self-driving car is that their cars meet the minimum specs for uber's program, which are probably more stringent than "must have 4 wheels" and yet still falls far short of "must have a lidar system accessible through an api".