I expect that at peak use (rush hour), the total number of cars needed would not be so substantially lower than the current number of cars owned. Perhaps a factor of 2 or 3.
Of course, with driverless cars, people may begin to stagger the starting / ending hours of the work day to allow for owning fewer cars.
But again, the consequence of a driverless car is that the steering wheel could be removed and fitting 4 strangers comfortably in a car becomes much easier. I would argue that carpooling/ridesharing would increase and ultimately decrease the number of cars needed.
That's true but consider that most of those people go to/from the same place.
If entity that provides those cars is big enough, they can match up the destination and optimize the traffic so that people going to the (roughly) same place share a car. When I was commuting from SF to MV for work, I had to use the whole car. There were plenty of other people who were commuting at the same time from/to very similar destination and used the whole car. It's not hard to algorithmically put several such people into the same car.
Today cars are 4/5 seaters so you get at 3x reduction compared to current levels (taking into account that not everyone drives by himself today) but you could easily redesign the cars to not be much larger and taking 8-10 people. Or make them straight up buses. Google already does that with their shuttles where they pick e.g. people from SF and drive them in big buses to MV, except it would be much more efficient because the potential pool of people transfered would not be just "people who work at Google and live in SF" but "every person who lives in SF and works somewhere in MV", which is a much bigger number.
As hard as it might be, it's a solved problem. FedEx is doing it. Uber is doing it. And I'm sure there are plenty of others out there, I just don't know about other industries.
"It's a solved problem"? I'm not sure what that even means. There are algorithms for finding solutions, and if you have enough time and storage space, you can solve it.
If I'm understanding the relevant WP articles correctly, the travelling salesman problem is not solved, since a solution for it would also imply a solution for all NP-complete problems. What these companies have therefore cannot be an absolute solution, but rather an optimal approximation which may or may not (but probably isn't, and probably can't be proved to be in polynomial time) equal to the absolute solution.
Of course, I don't have much firsthand knowledge of computational complexity myself, so someone else who know better should call me on any nonsense above. =D
Since you decided to nitpick: travelling salesman problem (i.e. finding an minimal route) is a solved problem as in "for decades we have known algorithms to find a minimal path and we teach them to CS grads".
What you refer to is the fact that the computational complexity (i.e. time to finish it) of known algorithms rises exponentially with the size of the problem and exponentially is a code word for "really, really quickly". It just takes too much time to find the minimal path if your graph is big.
What I meant, however, is that the problem has been solved in practice. When you ask Google how to drive from SF to NY, it'll give pretty good answer in milliseconds. Is it an optimal answer? It might be, it might be not, but it's a very good answer. Getting slightly better answer is not worth the computational time because it won't make a difference in practice in your trip.
Similarly, a car rental company doesn't have to schedule things optimally, they just have to schedule things really good, and that's possible with much less computationally expensive algorithms. The big win is when you go from "no optimization" to "good optimization", not from "good optimization" to "perfect optimization".
To be sure, I wasn't suggesting the problem hadn't been practically solved, and I understand quite well that beyond a certain point, there's very little incentive in improving the solution any further.
In retrospect, I probably should have kept my thoughts to myself, and I probably deserve the downvote or two that I got for not doing so. =)
> What I meant, however, is that the problem has been solved in practice. When you ask Google how to drive from SF to NY, it'll give pretty good answer in milliseconds. Is it an optimal answer? It might be, it might be not, but it's a very good answer. Getting slightly better answer is not worth the computational time because it won't make a difference in practice in your trip.
That's not the traveling salesman problem, it's the shortest path problem, for which polynomial-time algorithms are well-known and taught to first-year computer-science students. (To be fair, driving directions do require coming up with good edge weights on the graph composed of the American highway system, but once you've got weights, you can run Dijkstra's algorithm and you're done. That's probably not how GMaps driving directions work, but the point is that driving directions are not TSP.)
The post is misleading, people do not have multiple cars because they cannot drive, it is because they have to use it for multiple things at the same time. Morning going to work, dropping kids etc. will happen all at the same time. Peak car usage will not be affected by self driving cars.
I don't think peak usage is equal to the total number of owned cars; that would imply that there's some time of day at which every single car is simultaneously being driven, which is pretty hard to believe.
It'd require inter-household car sharing on a significant scale to take advantage of that, though.
So your hourly car rental rates will go up during peak usage until it balances out peak usage. While car rental companies may decline to change their rates on a minute-by-minute basis the only reason is to prevent customer confusion, nothing technical will stop it. Supply, demand, and price information can optimize a system like nothing else.
And even during peak times, car rentals will be augmented by a suddenly-reinvigorate public transport system; your car may take you to a bus, to be picked up by a car on the other end.
The only missing bit of tech is the actual self-driving car. Everything else is simple and obvious extrapolation from brain-dead simple Economics 101 and already-existing network technology. Modern car usage will be seen as only an incremental improvement on horse-drawn buggies, with the real revolution happening much later.
Actually, I think such a system would kill public transportation because it would become cheaper to use the cars than public transport.
I don't know the exact breakdown but my guess is that salaries of drivers are much more expensive than buses and gas, especially given those are often unionized jobs with very costly benefits (pension etc.).
That's a good point. I still think public transportation would be replaced.
I believe the main reason local government provides public transportation is because to achieve low prices it needs to be subsidized. No private enterprise is interested in competing in a market where you can't make profit.
At the same time, you don't see local government e.g. opening grocery shops: this need is well filled by private enterprise.
The car rental business would develop despite public transportation because it covers a much bigger needs. However, when it reaches scale and the low price that comes with it, it might make public transportation no longer needed as it'll also cover that use case.
Even if it's slightly more expensive, would you rather pay $2, wait 15mins, not get exactly where you want to go and hope no one will punch you in your face and steal your iphone (true story!) or pay $5, wait 5min and get exactly where you wanted to go?
If enough people start doing that, at some point the fixed cost of running public transportation will be too high to justify low usage by public.
Not to mention I do expect for profit enterprise to adopt driverless cars much faster than government agency, hence giving them a cost advantage for quite a few years.
People have multiple cars because they want to live among people who can afford multiple cars, and more specifically avoid the urban poor who cannot afford them. American car dependent suburbia exists in large part due to white flight. I imagine cheap driver-less cars would screw up this system.
[3] - Is all that matters. The industry and their constituents could easily push back driverless cars 20 years, and this is without any catastrophic accidents happening. So I'd say about 35 years before we see driverless cars disrupting anything.
Insurance companies matter a lot, too, though they don't get much coverage in the article. Like others, I don't find the arguments for lower overall usage very persuasive, but I could buy them being significantly safer.
If enough statistical evidence accumulates to demonstrate that these are safer than manually controlled cars, look for the insurance companies to impose a severe cash penalty on driving the car yourself. If the difference is large enough, you might find yourself unable to purchase insurance for a manually controlled vehicle at any price.
The movie iRobot was an interesting take on such a future.
In a few scenes in the movie, the female lead (the robot psychologist?) reacts in horror to Will Smith manually driving his petrol driven motobike because it has not auto-pilot and 'you realise petrol is explosive don't you?!?'
And after Will Smith has the forced accident in the tunnel (when attacked by the robots), his boss accuses him of unsafe driving because the blackbox reported that he had disengaged the auto-pilot.
So quite conceivably we'll end up in a future where manual car driving is seen in the same light as drink driving or speeding is seen today.
like any progressive technological development it will bring only improvements, including refusing to go to specific places specified by a transmission from central traffic control or automated blocking of doors and delievering you to court/jail/administrative hearing ...
There's another unintended effect I haven't seen talked about anywhere: when cars no longer require human drivers, the cost of driving in human terms (time, frustration, danger) will be drastically reduced, but the energy cost will only be reduced by a small amount. The logical consequence of this is that cars will be doing a lot more driving and as a society we will spend a lot more energy on transportation overall.
interesting link, do you think this increased use could go hand in hand with increased efficiency? think about it. if every car drove optimally, they would meet their planned MPG. Now no one gets the listed MPG as they all drive like, well humans.
The point here is that very efficient driverless cars may encourage more travel than driver cars do. The increase in total miles travelled may more than offset the savings for any given passenger mile.
The key to the Jevons paradox is that first order estimates of the effects of efficiency are misleading. You must also give regard to how the increased efficiency will change buyer preferences.
While I suspect you're right in that the # of miles driven would increase, I suspect that automated hypermiling would more than compensate for it, leading to a net decrease in energy usage.
I doubt it. How much more efficient do you think automated cars would be? Now think about all the people who will decide to live farther from work because they can work (or play!) on their commute, all the people who will take long overnight road trips while sleeping in the car, all the cars that will drive empty from place to place for various reasons, and most of all the businesses that will all have free automated delivery. I could easily see all that resulting in a 5-10x increase in car-miles driven.
This. A 3 hour commute would not be a big deal if you could sleep/get online/watch tv/do most of the stuff you would do at home. Commutes would also be shorter, all else being equal, since near-optimal automated driving would mean fewer traffic jams.
You'd find automated commuters living way out of town in areas that are currently nearly uninhabited. There probably aren't many places in the lower 48 that are too far from a medium-to-large city to commute with automated driving.
Road trips would also be easier - you'd have to be in a pretty big hurry to pay for an airline ticket when you can go 1500 miles or so in 24 hours with no effort, and without the need to pay for a hotel.
>you'd have to be in a pretty big hurry to pay for an airline ticket
That probably wouldn't result in a large net increase in total fuel consumption.
According to wikipedia the average commercial jet gets 49 passenger miles per gallon. It wouldn't be to hard to get that out of a driverless vehicle, add a second passenger and
you're twice as fuel efficient as an airplane.
Even if you get more passenger miles, there might still be an increase due to more total trips taken. Imagine if you could get off work on Friday, head out at 7 PM and make 600 miles by dawn while still getting a good night's sleep. You have the weekend to do what you want, and then Sunday night you sleep through the drive home, and are ready to go back to work Monday morning. I'd probably take a road trip twice a month. As it is, I wouldn't consider going 600 miles unless I had at least 5 days off, which essentially means once a year.
That is a possibility. But I imagine that when driverless cars become ubiquitous, most people won't own them.
If you did want to pay for a car to swing by your house to pick you up, it probably wouldn't be a car, but something more like a small bus with sleeping compartments that could pick up and drop off people along the way.
If you could fit 10 people along most of the route (and you could assuming it was between 2 moderately populated areas and road trips were as common as they probably would be), your passenger miles per gallon would be incredibly high.
Given that this doesn't happen today and that subtracting the cost of the driver isn't going to reduce the overall cost of such a trip significantly, I really don't see that happening. The biggest benefit for bus-like operations is not needing to have driver rest time, but lack of flexibility is still going to push people to their own private transport.
Combine driverless cars with maglev trains that can combine multiple cars into one long train with individual cars coming and going, and you may have something. I'm not holding my breath that I'll see anything like that in my lifetime. Transportation infrastructure in the US has gone nowhere in my nearly 40 years; I don't expect a great deal of change (to the infrastructure) in the next.
> subtracting the cost of the driver isn't going to reduce the overall cost of such a trip significantly
I'll calculate the costs 10 passenger van, if you could work it out so that it gets an average of 40 mph for say 20 hours out of the day. That's about 300,000 miles per year divided by 20 miles per gallon that's 15,000 gallons of fuel. At $4 dollars per gallon thats $60,000 per year in fuel costs.
I'll assume another $60,000 per year in maintenance and depreciation (this is probably high b/c you can buy a new large passenger van for under $30,000)
We're at $110,000 in variable costs so far.
Now at $30,000 per year per driver in total costs (could be less, could be more depending on area)
Driver works 40 hours per week with sick time and vacation time. You'll need around 5 drivers in order to keep the van operational 24 hours per day.
That's $150,000--more than half of the total costs.
There are also other factors to consider. Drivers, don't want to operate that far from home, so you have to pay more.
You have to set up a logistical solution so that you can refresh the driver every 6-8 hours. That means depots where you can exchange them.
This is already possible, until recently, my commute was 1.5 hours (each way) in which I would normally sleep or read, and occasionally watch TV. It wasn't a big deal, but I wouldn't have wanted to go much further, and there's no way I'd have gone twice that far.
It's still rare for people to commute that far on a daily basis. Everyone I know who travels more than about 2 hours from home to work tends to take weekday lodgings within walking distance of work.
A 3 hour commute will always be a big deal, because even if you work a strict 8 hour day, never staying late, that means that 14 hours out of 24 are spent away from home. If 8 of those remaining 10 are spent sleeping, you now have 2 hours left in which to indulge in any kind of leisure or childcare activity. Not everything can be done on the move - I can't imagine it being very likely that you'll be able to go for a swim or play rugby or cricket whilst travelling to and from work.
How much more efficient do you think automated cars would be?
Potentially a lot. In a 100% automated environment you don't need stop lights and traffic jams are substantially reduced. But yeah, I agree that the increased use resulting from greater convenience would outweigh that.
This is why I hope Google will license their technology only to all-electric cars. Both technologies are disruptive, and they might as well hit 2 birds with one stone.
Electric cars might not become the norm for another decade. The technology is still nascent and it will improve significantly over the next decade.
Also, why would you want to restrict an improvement that could benefit gas powered cars too? A self-driving car would probably be more efficient than a human driver. Fewer traffic jams would be less wasted fuel.
My assumption here is that when we have driverless cars, they wont be purchased by individuals but a few global providers will emerge that will provide those cars as a service (similar to how we only have few rental car companies, or how ZipCar is a leader in its field or how Uber is becoming a single provider of taxi-like service).
In that scenario, both the cost of cars and cost of energy (gas or electric) will have direct negative impact on their revenues so they will have a very strong economic incentive to drive down that cost.
A single buyer isn't well informed and isn't very rational, so a nice shape of the car might be more important than mpg mileage. Our hypothetical provider would, however, consider it one of the most important characteristics.
Combine that with a great purchasing power of such provider. If they buy 10000 cars and they want high mpg, you can be sure that car companies will spend most of their effort on increasing mpg.
Why wouldn't driverless cars be purchased by individuals?
I want one and I have money - when they're available, why wouldn't I be able to buy one?
Are you suggesting that I won't want to? I'm skeptical - the convenience of being able to use something whenever I want trumps cost. Or are you suggesting that I won't be able to?
He's suggesting that when you can open an app on your phone and (for the vast majority of the population in non rural areas) get a car to your door within 5 minutes, there won't be a connivence gap.
I believe the idea is that with time, it won't make fiscal sense. The first wave of these will be bought in small part by early adopting owners but more likely by small-time transit / freight companies looking to save a buck.
Once it's a proven small-scale model, larger transit and freight systems will start using them; I'd expect ZipCar and Uber and similar services to jump on it. At this time, more and more people would be buying their own driverless car.
Now picture 20 years down the road. You're a young person who can't afford a car. Do you get a job to pay to buy a car, or do you use a service like ZipCar? Fast forward 5 years, you've got a job and you could afford a car. Do you buy one to give you "added convenience"? That's what ZipCar will have to beat. If they can get a car to you by the time you're outside your house, they are actually beating personally owned vehicles, and the closer to that they get, the less likely it is people will purchase vehicles.
It's mostly because no-one will pay $5k extra for a car that they perceive to be a worse driver than themselves, but a better driver than a taxi driver.
There have been many studies that show that the large majority of people consider themselves "better than average" drivers. It is called Lake Woebegone effect or Illusory superiority - http://en.wikipedia.org/wiki/Lake_woebegone_effect
Certainly they do, but of course, they would think so. Still, I'd argue that this is not even that crazy. At least in San Francisco, I vastly prefer to ride with just about any colleague I've ever met than the local taxi drivers, based on these criteria:
- comfort while riding (lack of abrupt motions requiring me to brace myself)
- courtesy to other drivers and pedestrians
- actual number of traffic accidents (colleagues: 0, taxis: 2, although fortunately both accidents were mainly property damage rather than serious injury... but still)
I want one and I have money - when they're available, why wouldn't I be able to buy one?
I think you would, but the conventional wisdom around owning your own car would along the lines of the current CW surrounding owning your own aircraft: it would be something for wealthy enthusiasts, not the general public.
Planes are extra expensive because only coorporations buy them. There isn't infrastructure to sell them to individuals. That raises the price and difficulty of buying them, and also makes it a more sound political strategy to make stricter laws requiring airplane safety, which further makes it inconvenient to operate your own plane.
Some very rich people still own private jets, but it's rarer than it would be if more people were rich.
1. You have to change oil, tires and make all the other repairs.
2. You have to pay for insurance.
3. You have to pay for parking. In SF parking alone can cost you $100/month (I mean if you don't own a house with a garage (which will take you back $1 million); I don't think most rental apartments come with parking spaces so you either have to stress out about finding a parking space for the night close to where you live or pay for parking spot).
4. Parking while driving in the city. If you need to drive somewhere and stay there, you have to find parking, which, in SF, is not easy to do. A rented car just drives you to where you need to go and then goes of to pick up the next customer. When you want to get back, you summon another car.
5. What if you want go skiing for the weekend and need snow tires? Would you rather put snow tires yourself or summon a car that has them?
I'm in SF and I already use my car so rarely that its battery died on me. I do ZipCar but it doesn't cover a lot of scenarios that, while not frequent, do happen and the requirement that you need to return the car to where you took it from makes it non starter for short trips within city.
If there was a service where I could just summon a car anytime, it would be cheap enough to work as a taxi replacement but also for longer trips, I wouldn't buy a car.
Cars are much more than solutions to a transport problem. They are status symbols, personal identities, an extension of ones castle. It will take a very long time, perhaps never, for most people to adopt a taxi/public transport route.
I'd very happily own a driverless car. Even better will be when I take it out of driverless mode in the rural roads. It would be able to take itself to get its oil/ires serviced,to be recharged and to park somewhere cheap and out of the way. Bring it on.
Cars are status symbols in large part because they are advertised as such. If people buy fewer cars you will see a lot less car advertising which will directly reduce the status symbol effect.
According to the wikipedia article, if we decide that the decide that increased energy (or road space etc.) use is a bad thing we can fix Jevon's paradox with taxation on car use.
Road space shouldn't be a problem, from studies I've read with increased speed and traffic density that come from driverless cars you could fit several times more traffic on the current roads.
If every car is driven sutomatically, cars can be driven much more efficiently, since there won't be any need for stop signs or traffic lights in the current form. Cars can just coordinate that one car slows down a little and one car speeds up a little. The only reasonfor card to stop will be pedestrians.
Of course I assume that there will be only driverless cars. I am fairly sure that human driven cars will be illegal in less than 20 years. Most accidents will be caused by human driven cats which will lead to a ban. I also predict that the US will be one of the last western countries to ban human driven cars
While this concept sounds amazing to me, lets not forget that we are not the majority. I just got off the phone with a relative who hates taking a plane, they would rather drive 24 hours to their destination. I consider this concept to be pure insanity, but i don't think this is limited to a small portion of the populous. There is much more at play in the mind of the 'average American(in this case)' than efficiency and 'the car will drive for me'. The lack of a feeling of person control is not among the lesser of these issues.
i don't think this is limited to a small portion of the populous.
I do that.
But it's not because I hate planes but I've got three kids. Airfare for five to visit grandma in Virginia is prohibitive compared to the cost of driving.
I'm not a nut about it: if it's just my wife and I, we fly.
But on those long drives especially, I'd love a car to do the driving.
I like driving and make no apology for it. I love to set off and drive for a 1000 miles or more. I enjoy driving to no particular destination in particular. It's no more or less rational than liking golf or playing computer games.
Actually the energy used would be reduced by quite a bit. They can drive in formation. This reduces energy expended by reducing drag. It also reduces congestion. It is estimated that congestion wastes 2.4B gallons of gasoline in the US [1]. They can take alternate routes. And if we get smart traffic signals, then they wont have to stop as often by coordinating speed with the flow system. Not to mention eliminating people flooring it out of frustration.
Given the current (and undoubtedly soon aggravating) energy crisis, on the contrary energy cost will probably be the limiting factor in most human activities in the coming decades, therefore shared cars will be making even more sense.
the flip side is that if they might do better with constant use. Diesel engines for instance, do better under constant, consistent load. Engines would remain lubricated and heated. Stratification and temp changes would have less effect, etc.
Are you sure about that? If it's the norm to pay for a share in a fleet of cars, then those cars will probably be owned and managed by a company that will ensure that they get timely preventive maintenance. Yes, there will be changes in how much is reasonable to pay for added durability, but do you have any evidence that cars need to be radically redesigned?
What's that got to do with anything? Why should driverless cars be expected to last as many years as current cars? If you're only paying for 20% of the car's lifetime cost, then you're not going to mind if it has to be replaced five times more often. In fact, you'll probably like it - that's why anybody ever leases cars.
It's quite reasonable to expect that it will be economical to pay something like 20% more for the drivetrain to make it more durable, but that's not a radical re-design.
sending your car home for your spouse to use and then back again to pick you up at the end of your day means it's using twice as much fuel on wasted trips with no occupants.
Well by the time we get driveless cars, we'll probably be on the tenth generation of 'Prius' technology and the wasted fuel cost will be marginal compared to the cost of a car.
Using a 2000 pound block of steel to transport one human being at high rates of speed is fundamentally energy intensive. Even with rainbows and pixie dust it would waste lots of energy.
As the article stated, the automated feature reduces the need for multiple cars, it enables better car pooling options - so yes, if you want to look at it by narrowly isolating one aspect of the scenario, sure, it uses more fuel to drive a car anywhere.
But as a complete solution - which is what the article talks about - the cost of 'wasted' fuel is negligible given the change in behaviours and travel methods that automated cars may bring about.
You arent seeing the bigger picture here. If you choose this level of efficiency, after getting dropped off at work, 'your car' will simply locate the next closest customer. If you want to spend the resources on your own private car or a more available one, more power to you.
But people tend to live in the suburbs and work downtown; so the net traffic in one direction is much greater than the net traffic in the other direction.
I don't think the article's argument justifies the idea that fewer cars will be sold.
Cars wear out primarily through driving. If you simply switched to automatic driving, what you would have is fewer cars being driven more often and so being worn more quickly. If car-mile consumption stays the same, new-car production would stay the same.
On the other hand, if the auto-drive cars increased carpooling, then you'd see a decrease in car-miles consumed and so a decrease in production. But if auto-drive cars drove around empty more, you might have even more car-miles being consumed.
Moreover, you'd have a "big bang" where people decide to mostly stop driving the old, non-automatically-driving cars and so there'd a huge spike in consumption at that point.
The space saved by avoiding parking could be really large, still.
A nice thing would be that at the start, a person might be able to finance their self-driving car by renting in out when they didn't need it. Those economies might make the phenomena spread really quickly.
If you simply switched to automatic driving, what you would have is fewer cars being driven more often and so being worn more quickly
At the very least, I expect that would be offset by some amount by the fact that automatic cars are going to be driven in a much more efficient and car-friendly manner than they are when a human is operating it with their imperfect motor skills, their ego, their laziness, etc. The way the cars operate would be (or at least could be) algorithmically optimized so that all those little things you're encouraged to do (like pretend there is a glass of water on the dash when you're driving, etc.), actually happen.
Also, once Jiffy Lube starts accepting driverless cars, you would see oil changes and maintenance done on a more regular basis since you wouldn't have to drive it to the shop, further extending the lifespan of the vehicles.
Oil changes can easily be automated. Electric vehicles? All they need is a parking spot to automatically pull into and charge at until required to move to their next waypoint.
The passenger-mile invariance would be offset by the selection of different design tradeoffs. For example, most driver cars have far more performance "on tap" than will be used in the entire lifetime of the car.
Such performance is not free. It causes engineering tradeoffs that increase maintenance costs, complexity and of course reduced fuel efficiency.
But it's easier to sell a driver car with some oomph than not.
There will be other design tradeoffs you and I have not thought of (and could not think of) that will add up to further efficiencies.
The article might not but I think that it would totally happen.
My assumption here is that there would emerge a big car provider who would provide those cars, similar to how there's only few major car rentals, or how ZipCar is leading in its category or even how Uber emerges as a single taxi-like provider. The economies of scale are very big here.
Given that, some things you have not accounted for:
Higher average utilization per car. I don't know how much commute traffic accounts for total traffic, but by judging rush traffic on 580, it's quite a lot. Most of those people go from the same location (e.g. SF) to the same location (e.g. Mountain View) and today most of them drive alone. Today 4 or 5 of them would be scheduled by an algorithm to use the same car.
Because of that the cars would get re-designed for a higher capacity (at least 8-10 people) or even straight up buses. A big provider would have so much data that they could optimize the hell out of sharing cars.
Similarly, the cars would utilized, on average, much better. Again, a company with enough data could really optimize for a lifetime of a car and they would have economic incentives to do so (unlike a single customer for whom even gathering the necessary data would be cost prohibitive).
Our cars are, without a question, utilized very inefficiently. Someone operating a fleet of tens of thousands of cars would have not only economic incentive but also necessary data and necessary transaction volume to optimize per-car utilization, because for them it would translate directly into large savings.
I can't upvote this enough, people seem stuck in their concept of what a car is. Without a driver we are free to redesign cars in so many ways.
We no longer need to worry about designing it so that the driver can see every angle, so we can design more efficient seating layouts. We can even add partitions so the 10 passengers don't have to see each other.
Long distance deliver vehicles can operate at lower fuel efficient speeds.
In fact, many transit systems do combine delivery and passenger vehicles -- greyhound has a package delivery service.
To your other point: I think that assuming we'll go off the wall and design wacky new vehicles is too optimistic, akin to the "cities of the future" envisioned during the past hundred years. People's sensibilities of "what makes a car look nice" and "what is a metal box I'd like to sit in" seem to be pretty set, and we don't have a very different experience in a train or on a bus today than in a car. Face forward or backward in a seat with big windows and high visibility.
Part of that is for your senses to not freak out, part of it is just that it's cheap and easy, and part of it is that experiments in the form-factor don't need to be done, so they aren't done. I expect the same trend will continue unless flukes occur.
>In fact, many transit systems do combine delivery and passenger vehicles -- greyhound has a package delivery service.
I didn't know that about greyhound. I'm aware that airliners routinely carry packages. I was thinking more, you've got a car coming to your house every morning, let's throw your Amazon.com orders on it.
>design wacky new vehicles
I'm not thinking wacky designs. A car is usually designed to be sat in for a certain length of time. They aren't really designed for very long trips without stopping frequently, since driverless cars make long trips much more appealing, they'll have to be redesigned to an extent.
A fleet of rental cars would be much more efficient if each car held say 10 people instead of 4, so another natural redesign.
> "what makes a car look nice"
To many people, a car becomes part of their identity, not so much with a temporary rental, so I think function starts to trump form
>don't have a very different experience in a train or on a bus today than in a car
I think the experience in a long distance train is very different than in a car.
The main difference is compartments, and that's what I envision for cars. I'm not talking anything radical, just what I think would be natural progressions when you remove the need for a driver.
A 10 person vehicle necessarily needs to be larger and have a more powerful drivetrain so it would be less efficient in the dead miles it has to drive to get around and pick people up. So I don't see that 10-person vehicles would become the norm. I would actually see it going the other way, and that 2 person vehicles the size of a smart would actually become the norm. Most people currently have excess seating capacity (5) for the few occasions when they need it. But with driverless cars, you don't need that, because you can spread the load over several cars. Why drive a 10 capacity car around for the odd case where 10 people need to go to the one location, rather than 5 2 person cars that can assemble when needed to cart 10 people on the same trip.
For peak times, I'm imagining 8-12 seater vehicles where people sit in pods facing outwards, with dividers that could be retracted if you were speaking with your partner in the same vehicle. Pick-up points would be very near to your home, and drop-off very near to your work. No parking to worry about. Something in between a regular driven commute and walking three blocks to wait for a bus.
Software would decide routes and timing to best service as many people as possible.
While I'm all for your point that drivelessness will allow an incredible re-imagining of the automobile, considering the gp responded my original post, I just want to bring things back to the question of resource utilization.
The automobile today is half efficient transport and half personal expression/personal entertainment. And the personal expression/personal entertainment part is where the massive resource utilization comes in. So, sure, you could reimagine the automobile for super-efficiency with four-people per car whenever you're driving and the resource utilization goes away - so does the personal entertainment/personal expression stuff. So you could go multiple ways. Towards a super-efficient taxi and towards an office/living room on wheels. The first way would involve less resources consumed, the second would involve less resources consumed. It is hard to be certain what the net outcome will be.
I think the natural progression of driverless cars is away from individual ownership. Why own a car when you can have near instant availability of a much cheaper rental.
To many people, a car becomes part of their identity, or like you said
>expression/personal entertainment
I'm sure that there will still be people for whom this is true, but for most of us it won't matter b/c the efficient always there taxi will be so much cheaper.
For most middle to lower-middle class Americans I think current cars are really out of their comfortable price range, they own them b/c they view them as a requirement.
If you give them an alternative that is just as convient, but cheaper, and without the maintenance hassles, it's no contest.
Good point about efficiency gains becoming a more direct incentive; I look forward to it. However, I doubt that optimization would trend solely toward high capacity vehicles - think of taxis airport taxis as two classes of hireable fare, divided into single user single destination and multiple user single destination. What we need is a multiple user multiple destination system.
Taxis are for in-city travel and occasionally nearby inter-city travel, and the regular car has ended up being fine. They do have a few larger vehicles for big groups, but those are special case vehicles. On average, they tend to get 1-2 people at a time wanting to go places.
Airport taxis tend to be closer to a commute-oriented vehicle. Several people are going from here to somewhere else, so they tend toward 5-6 person vehicles which operate through the night on >1hour long trips.
Corporate transit systems like Microsoft's Connect system are essentially multi-user multi-destination. Those make buses make sense, but there are a relatively small number of buses making set route transfers while a hive of taxis provides last-mile service.
The "Future of Transportation" article from a few days ago [1] suggests that the average vehicle has 1.2 people in it, and that apparently takes commuters into account. Now we must consider: will behaviour change to increase carpooling?
Specifically, what is presently preventing people more people from taking buses to work? (lots do take buses and trains already, but many people drive 10 to 100 minutes)
It could be that a bus doesn't get you /right/ to work. It could be that you would never think to contact everyone in your neighbourhood and figure out that you could buy a bus and run it just for you guys. It could be that the cost of owning and using a car is just low enough that it doesn't really save you anything. It could be an aesthetic sense of independence.
My bet is that this will work out like Microsoft's system. Major transit routes can scale passenger levels up and move efficiently between well placed bus-station/pickup hubs, and armies of taxis will shuttle you to your final destination. Google's bikes are nice, but last-mile service will have to deal with a broader sprawl. Most of them will be focused on getting from hub to delivery and vice versa during morning and evening, but then, what happens during the day? Even weirder: what happens at night?
Unfortunately there are a few assumptions being made here that don't really jive with car usage in practice. 96% usage? People's schedules aren't that flexible, my spouse and I commute at the same time to different places, and a large chunk of the people in my apartment do as well. This is known as rush hour.
Not to mention that driving places, without actually driving people places (ie, back home to drive the wife) effectively doubles the mileage, and thus my gas (or the energy source du jour's) consumption.
edit: Also, to build on the above comment, the benefit of having a car is having the freedom to use it when you want, it's always there. Driverless cars always in motion are basically public transit. Just take the bus/taxi/subway already.
>Driverless cars always in motion are basically public transit
Minus the cost of the driver.
The point of ubiquitous driverless cars, is that one could get to you so quickly that for all intents and purposes it would always be there.
Also they don't always have to be in motion. Just give people a discount on service if they let cars park in their driveways while waiting for instructions.
It is a tragic failure of imagination to believe that a driverless car will offer the same opportunities as a bus or subway. Yes, I'm sure there will be a public transit version of them. Yes, I'm sure that you'll occassionally have to reject one because there's a pool of vomit it in it. But, no, it'll be nothing like walking down the street for a bus, or, for a young woman, hailing a taxi driven by a stranger, at night.
Assuming usage does not go from 4% to 96% and he is only partially right, will there still be no interesting societal transformations along his outline due to an increase to, say, 50%?
My first thought about driverless cars is that stock in alcohol companies would skyrocket. I know a good many people who would go to the bar on a more regular basis and now that they have a driver for the ride home. Also on a week day after work, why not have another beer? No risk of hurting someone or a DUI, right?
America needs to get back to its roots as the premier boozing nation. We were long known for this. Hell, there was practically a civil war over the whole whiskey rebellion thing.
Different jobs will take their place. There are a lot fewer farmers these days, for example. The real trick is to find effective ways for people to get the training.
if somebody has worked most/all of their life in a low-education unskilled labor job (the kind that are most likely to be disrupted by computers), who says they even have the ability to become a robotologist or something very intensively knowledge based? In the future these will really be the only kinds of jobs computers cant do. The problems here are much more complex than the Luddite issues of yesteryear, and its disingenuous to write somebody off as a dumb Luddite.
While technology makes easy jobs disappear, it also makes hard jobs easier. The advent of calculators means grocery clerks can be completely innumerate.
The Luddite argument is going to become obsolete one day. There are only so many iGadgets and squeaky plastic toys people are able to pay for. Just because an argument has been correct for along time, doesn't mean it will always be correct.
This is the wrong place to make this argument (it should probably be made at length) but I think the only real solution when we have a super-efficient, highly automated world is to drastically increase taxes to the super-rich and subsidize people with obsolete skillsets. But regardless, we'll no doubt see more interesting developments here in the next ten years. The current unemployment situation is just the beginning, although a lot is going to happen in the meantime.
(This is from a country where about 10% of the population is currently on welfare and we almost de-facto already have the economic subsity system I'm describing)
We're in the software business. Putting people out of jobs is what we do. Pretty much any technological advancement puts people out of jobs. But it's okay! There are more productive things they can be doing. That's kind of the whole point!
Kevin Drum has argued that once computers are smart enough to drive trucks, what are truck drivers going to do instead? Computers will be smart enough to do just about anything else they could do.
As the efficiency with which a resource is used increases, the tendency is to use more of it rather than less.
Concretely, if I don't have to drive the car myself, I wouldn't hesitate to drive anywhere. I'd go to the city every day if I could read on the way, and sleep on the way home.
With shared resources, this might not be the case.
Today, car owners have already spent $x on their cars -- so they have an incentive to use their investment to the fullest.
Compare that with a shared-car system like ZipCar. ZipCar is generally much more efficient than owning a vehicle (if you don't use it to commute) -- however, since you pay per hour, you have an incentive to skip the car when it's not necessary.
By removing the sunk-costs of ownership, you could incentivize conservation rather than consumption.
With ZipCar, you have to first get to the car, and then also get back from the car. For most people, the car they own is in their driveway or parking spot.
With a model more like Car2Go (pick up the car wherever you can find one, leave it wherever you want to) this isn't as much of an issue. In Austin there's pretty much always a car within a few blocks... almost "driveway-level" accessibility.
And if even 10% of drivers switch to using this model, then there's always going to be a car on your block. That's pretty competitive.
With driverless cars it's even less of an issue, because the car will literally drive from the end of the block (where it's waiting) to your door.
Assuming, that the San Diego prices can be applied to Francisco, it seems like taking such a car from downtown to the airport would be about the same as the BART subway.
I assume, that would be too expensive for a daily commute.
Maybe - I wonder just how elastic the miles-traveled vs convenience curve is.
There is surely an upper bound to how much traveling any person wants to do. If it were 5% more convenient for me to travel by car, I _might_ travel 5% more, but I couldn't see myself traveling 50% or 100% more no matter _how_ convenient you could make that travel.
This is a far down the road vision. It calls for a huge shift in the mindset of the consumer AND the manufacturer, to say nothing of the society. Then there is the infrastructure and error handling - what happens when it gets a flat tire on its way to pick up the kids after dropping you off?
Lets give it 15 years. 15 years is a long time for technology but not so much for societal evolution.
I can see the driverless car being on the road in 5 years and I see it being used for carpools. But the vision as painted in the post? Yes, I'm a skeptic.
Very interesting stuff though...I can't wait till they hit the road.
Who is liable if your driverless Civic drives you into a tree? What if it runs over a pedestrian? Honda? Google? You?
Driverless cars could dramatically reduce the dangers of driving, but it's a lot more complicated if a machine learning algorithm kills somebody than if a mistaken human being does.
All it needs is for someone to take liability. If you own the driverless car, then you buy the insurance. If it's leased from a pool, then they buy the insurance.
The software vendor for their part take an over-arching insurance on software errors across thousands of cars. Done deal.
Today if you have a 20 year old your premium is higher. Tomorrow if you have a driverless car, your premium might be low even if you have a 16 yr old. Parents buying a first car might just buy a driverless car for a lower insurance premium & the safety of not having the kid drive!
It seems that to most Americans, the draw of the driverless car won't be "hey, this is more efficient and is costing me less overall." Rather, I think it will be, "hey I don't have to pay attention in the car anymore, I'll be on my phone playing on the internet the entire way home!!!!!!!" Followed shortly thereafter by, "hey why doesn't this car have the internet in the front seat?"
According to the U.S. Department of Transportation, Americans spend more than 500 million 'commuter hours' per week in their automobiles. Once driverless cars take over this time can be spent on browsing web (clicking ads)
I brought this up in Matt Maroon's post about the topic back in January, and he brought up a good point: People like to own their car because it's a moving storage locker.
The use case I'm imagining is people going shopping, and wanting to put bags from one store away in their car before going into another. So what if you could summon up a self-driving car, put your stuff in its trunk for a fee, and then call back the same car later with your stuff still in its trunk? It could go be a taxi while you're busy shopping, and not allow other customers to unlock its trunk.
But I think the removable trunk will be combined with a passenger-less unit that carries the trunk back to your house and drops it off. Sort of a tiny version of container shipping.
That sounds like something that might happen in the farther future, but I expect there will be a lot of general-purpose auto-driving cars made before companies start making super-specialized models like that.
This is probably one of my favorite things about owning a car, and why I have a sedan vs. a hatchback. Being able to store emergency supplies (couple cases of water, a backpack, ...), personal tech supplies (spare mbp charger, cable toolkit, etc.) in my car makes life much easier. Realistically this is 100-200 pounds of stuff, and I wouldn't want to have to carry it to/from a car myself every time I got a robot car.
So the effect of a driverless car is going to be a driverless taxi. Why doesn't everyone use a taxi now? I'm sure it would cost less than cars do. But nobody really does the cost-benefit analysis of these sorts of things.
Eh, that's doubtful actually. In most US cities (other than, say NYC, SF, and a handful of others), taxi service is expensive and slow to arrive. Parking is free, and cars can be obtained relatively cheaply.
Taxis today aren't driverless, and paying a human driver is a significant cost, especially in any country with a relatively high cost of living (and therefore high labor costs). Driverless taxis could change this significantly.
Lots of people do already use mass transit every day, which has a cost structure closer to driverless taxis (because the labor cost of a bus or train driver is amortized across more passengers).
"And if cars are receiving 20 times more actual use, that would imply that there would be 20 times less cars sold."
Incorrect conclusion. Cars being used 20 times more will break down 20 times faster. Less might be needed on the road at the same time but they'll break down and need to be replaced at a much faster rate. Over a serval year timeframe the amount of cars turned over will remain roughly the same.
I also wouldn't send my car home to pick up my wife and drive her to work because my fuel bills will skyrocket through all the extra driving around.
The shared resource makes sense. It'll be similar to taxi services that exist now. You call for a driver and the closest shared car will come to pick you up.
>Cars being used 20 times more will break down 20 times faster.
Not true if they're owned by big players who professionally maintain them. Want to buy a used LLV (postal truck from the eighties/nineties)? You can't, because they haven't been surplused yet.
Also, cold engines wear much faster than hot engines, so an engine that runs all day does not wear out as quickly per mile as one that's driven a few times daily.
Sure, they haven't been surprised but they surely have had enough parts replaced that we can consider then new cars several times over.
You make a good point that a fleets maintained better than the average citizen but they will still break down and still need to be replaced faster that the author is stating. We will still have roughly the same amount of turnover over the course of a decade.
I'm not sure if anyone else has realized it yet, but driverless cars will have to be taxed on a time-on-road basis, rather than at a flat rate.
The reason? The economics of electric cars and parking.
Initially it seems that they will fix the parking problem - driverless cars can be sent to park outside the CBD, reducing traffic and freeing up space.
But the problem is that people/software will optimize for price, and for electric cars the cheapest scenario is for them to be stuck in a traffic jam on a public road.
Instead of going to a parking bay, the software will route them to the nearest traffic jam, where the car can sit with the electric motors off for a large amount of time. Inevitably, some software will misjudge how long their charge will last, their batteries will run flat and the traffic jam will get worse.
As far as I can see the only way around this is to increase the cost of being on the road.
Who exactly will do this? No major car company would risk their reputation selling a car with that programmed into its software.
What's the risk, exactly?
Our car has an "ultra-economy mode" where the motor turns off when stationary - many cars already have this.
Our car can park itself - many cars already have this.
Our car will find a parking space, optimizing the route to and from the parking garage - that seems logical, why would a company not do that?
Our company will optimize the route to and from the parking garage, and the time spent in the garage vs on the road based on price vs economy of the car - why would a company not do that?
The problems only occurs when hundreds of cars independently make their own decision to drive as slowly as possible to and from the parking space.
(Edit: someone else pointed out that UPS routes their trucks to avoid left turns. This isn't too dissimilar to that - they do it to improve fuel economy, this would be done to improve costs too)
If you made it a flat fee, then it's kind of like runs-batted-in. The car has to get more people delivered to make money, so once it gets you to where you need to go as efficiently and quickly as possible, you get out and it speeds off to complete another fare.
Increasing road-cost would have the opposite effect, I think. If the cars make money just by being on the road and not by completing tasks, they lose the incentive to finish and jam-sitting becomes optimal.
But I'm not optimistic about the share-car thing, at least in cities without a strong existing public transport system. In most cities too many people travel at rush-hour, with unique route requirements (dropping kids at school etc).
Why are there going to be traffic jams? Computers are driving the cars. There's not going to be accidents anymore. There's not going to be that weird bottleneck effect where one sheepish person slows down to much and eventually everyone behind them stops.
There will still be traffic lights, bottlenecks, and for the foreseeable future there will still be conventional cars.
Driverless cars will still be limited by physics (their size and how fast they can accelerate). If car routing algorithms deliberately route cars to places where the traffic is slow (because they can loiter, saving on parking while using little power) then the traffic jams will be worse.
If we rebuilt the road system specifically for driverless cars maybe we could eliminate traffic jams. Otherwise...
In such a case, parking costs will reduce. Since the number of cars are expected to be lower if they are driverless, the amount of space required for parking will be less. It would also mean that expenditure on the prime real estate required would be less (marginal costs are higher, marginal decrease would hence be more benefitial) and that will result in cheaper parking.
Secondly, it also opens up the scope for private taxies. A person can allow his car to be utilized by a certain section of his peers - colleagues, neighbours etc for low prices - resulting in lower parking time.
It'd also be possible to double-park (triple, or more even), using less parking space. If the car in the middle needs to get out it can ask its neighbours (several levels deep) to shuffle around and make a path
Traffic lights have no purpose except for conventional cars. Driverless cars can just go through intersections whenever no vehicle needs to go through perpendicularly. All intersections would just work like four-way stops, minus the requirement to stop when not yielding to another vehicle, and with more batching and parallelism for efficiency rather than strict serialization.
Fuel taxes already do this, while providing a strong incentive to run smaller and / or more efficient cars.
Roughly, 50% of my fuel bill is on direct taxation. That's it's primary tax cost, not the fixed annual license cost, and gives me a tax bill that's elastic by usage in exactly that way while also promoting efficiency.
Of course, best wishes to anyone trying persuade the average man in the street that this is A Good Thing....
They can turn off their motors without losing any efficiency (which is why hybrids do so well on efficiency in city traffic). Depending on how slowly the traffic moves there will be a point where it is cheaper to stay in traffic (with the motor off x% of the time) compared to paying for parking.
Why the assumption that electricity for vehicular use will be taxed at the same rate as for domestic use when there's wide-scale adoption of electric cars? Red diesel, for agricultural use off-road, certainly isn't taxed at the same level as road use diesel in the UK.
Interesting post. However, I think the primary consequence of driverless cars could be more mundane - everyone will end up using cabs/taxis far more often. This is the life I lived in Bombay a decade ago. The parking, cost overhead of a car is so high and rickshaws/cabs are so cheaply and plentifully available, that walking out of your home and hailing a cab to go wherever you want is the best option.
Cabs in the West are expensive because drivers need to be paid more and cartels are at work. If both these costs are eliminated then what works in a country like India is, suddenly, the best option for commuters in the West. Realtime pooling of cab passengers is a possibility, a traveler could be offered a choice of picking a co-traveler for a reduction in fare(or even cash back!) when already travelling in the car.
Families could still own cars, but they will probably pay for insurance on a per-mile basis, rather than a per-month basis, so that the incremental cost of owning a car is minimal. Effectively, you end up owning a cab that only you use and pay cab fare to the insurance company :-) The ratio of cabs to private cars would then be an interplay of insurance costs/parking costs and ride sharing benefits, with cities leaning towards cabs to escape parking costs. So, things wouldn't be too different :-)
One unfortunate consequence could be the increase in suburban sprawl and traffic congestion with driverless cars as people start caring lesser about longer commute times. Automated cars will probably be better behaved in traffic, but the road capacity will be pushed to its limits.
Most traffic is caused by the delay in human reaction times, and driverless cars can operate safely at much higher speeds and at much higher densities. Roads could handle massively more traffic if they were only packed with driverless cars.
Even if this is so, a computer-driven cars can certainly use normal junctions much more effectively than humans. Robots could do this: http://www.youtube.com/watch?v=Vnba7Y86jC4 [lucky driver goes full speed through dense traffic] on routine basis.
If you can just network the driverless cars up to the existing public transportation systems then we'd have a truly useful device.
Just walk outside your house and a car drives on up, opens its doors, and takes you to the most convenient location to transfer to a public bus, or if economical to pick up other passengers going near your destination.
And if cars are receiving 20 times more actual use, that would imply that there would be 20 times less cars sold
Actually, no it wouldn't. If cars are getting 20 times more use they will wear out much more quickly than they do now. That means cars will have to be replaced much more frequently. There would be fewer cars sold than there are now, but it wouldn't be 20x fewer.
The operating percent of a car will go from 4% to that 96%
This seems wildly optimistic to me. The driverless cars may be capable of driving around 96% of the time, but that doesn't mean they can be carrying people 96% of the time. No matter how efficient the system, if there are enough cars to handle peak traffic during the day, then a lot of those cars will be sitting around doing nothing at night.
I wonder what sort of population density and usage pattern you'd need before that worked in your favour?
If my apartment complex had a few dozen cars shared between a few hundred apartments, perhaps the car I take to go shopping could pick up the next door kid from soccer on the way back, then a different car might pick me up at the shops when I'm done, after dropping some other neighbour at the movies...
(I guess I'm now describing taxis. I wonder what the difference between this, and a taxi network of driverless cars is?)
Anybody got both Travis Kalanick and Elon Musk's numbers in their speed dial?
What if a "disrupt the cab industry" company got together with a "low moving part, high reliability electric car maker" to do an end-run around the expected auto industry opposition...
A fleet of driverless electric taxis, all routed by smartphone apps and behavioural prediction...
Then, when everybody is impressed with how well they work, you start selling fleets of them to Apple/Google/Oracle/SouthBayTechFirmDeJour - every evening a train of autonomous cars starts arriving and emptying out your campus 4 people per car heading for nearby/on-the-way destinations, all of a sudden those 20 hectares of parking lot can become cube farms or data centres...
During the middle of the day and all night, you lease the capacity to FedEx or UPS...
As long as we're hitting this one out the ballpark we can imagine all the online services getting in on the action: like the OKCupid speed date commute, the Yelp surprise me whats for dinner restaurant ride, or the Groupon deal of the day carpool.
Limited range cars with long recharge times would be a poor operating fit for economics that favor high utilization. There'd be too much downtime during peak times of day.
Maybe it could be coupled with swappable battery pack stations that the autotaxis could visit to get a freshly charged pack. A geographically focused taxi company could have the financial resources to invest in the battery depot, which would also solve the standardization problem.
Sorry, that's been solved. There are efforts to implementing this exact network of fuelling stations (battery swapping) for electric cars, and Better Place is a shining example from Isreal:
On the other hand, if you find yourself running low on juice, either because you're on a long trip or because you didn't plan ahead, it's not a problem. You just swap cars like the Pony Express swapped horses.
Lower utilization is balanced by lower energy costs than gasoline (but you're right, swappable batteries would help too).
And I assume you will install and deinstall the carseat each time? A huge portion of cars are used to transport kids, and it's not practical to remove the carseat each time.
You can't have a car with a loaner carseats either since they have to be individually adjusted to the kid.
Every single time people talk about cars, and public transportation they always forget about kids. I see it over and over. Come on people - expand your worldview a little.
I would expect that eventually the safety of automated cars will far exceed current levels and therefore child car seats won't be required in the same way that no-one uses a car seat on a train (or even a bus actually, which is presumably much more dangerous than a train)
Considering that there are 5 different types of car seats, and they can be adjusted to about 10 different configurations that's not going to be an easy task.
The people who design car seats these days know that their customers would rather spend an hour rethreading/reinstalling their car seat every six months than spend over $200 on the seat. But if there were a large fleet of shared cars on the road, and if some entrepreneur came out with a $1000 car seat that could be readjusted in seconds for a child of any size, then the owners of the fleet would have an incentive to loan out those seats along with the cars.
This just opens the market up to build a carseat that solves this problem. How about a regular seat that can be folded out or transformed into a seat suitable for children?
Where I live, a certain percentage of taxis have big trunks, for people with baggage (they can request that when they call for it).
Why not have a percentage of driverless cars pre-equipped with carseats? You'd just have to configure once the age of your kids, then when you call a car (using e.g. a smartphone app), you just say for who it is ("Siri, I need a car for john, mary and me") and the right type will come.
The car seat needs to be individually adjusted, especially for younger kids (different heights needs the straps in different slots, although the fine tuning can often be done on the spot).
And you're going to need every combination of ages, people have more than one child.
There are approximately 10 different car seat configurations for the various age ranges (5 actual car seats). And assuming up to 3 kids would require 1000 different cars.
There is rear facing (with 3 heights), front facing strapped (4 heights), front facing buckled, booster, and booster without back. (Although many car seats can handle 2 types in one seat. But it requires you to reinstall it.)
I'm not sure how it applies to other countries, but here at least you have to have special car seats for all children until they're aged 12. It's insane, but that's the fabulous new law they enacted. So, yeah in some cases, some parents really are buying that many seats.
They buy 3 since virtually all can convert between two levels (any two adjacent levels). A rare (and expensive and heavy) few can do 3 levels in one.
But converting the seat requires rethreading the straps and other adjustments - it can take an hour to install some types if you are not familiar with it.
If cars are getting 20 times more use they will wear out much more quickly than they do now.
Existing cars, if driven 20 times as much, would wear out 20 times more quickly, like taxicabs do now.
But it's also possible that cars would simply be built with more reliable components and more durable materials, like current aircraft and public transit vehicles are.
It's not cost-effective to build an ultra-reliable car that's sitting idle 96% of the time, but the economics would surely change if the utilization rate is much higher.
The thing is, taxis _don't_ wear out as fast as you'd think.
I remember talking to an old-school cab driver a while back, when I noticed his odo had ~650,000km (~400k miles) on it. We chatted a bit about it, and when I asked "So how long do cabs last" he said "3 good crashes." It doesn't which of the locally popular cab models you buy and it doesn't matter how far you drive them - you might need to fit a reconditioned diff or gearbox or even motor, but all of that is "routine maintenance" from his point of view. Its after the third time you've crunched it hard into something that it's time to get rid of it...
Yeah, but if you replace the diff, gearbox, and motor, how much of the cab is really left (mechanically speaking)? Replacing the engine is, from an engine-maker's perspective, the same as buying a new car.
Yeah, but from the cab owners perspective, $600 for a reconditioned diff or $1200 for a reco motor is a lot less hassle/expense than $40k for a new car, then getting all the cab-specific fitout done to it...
(Yeah, that's what the base model cars used for cabs cost here. Google Holden Commodore or Ford Falcon prices. New car prices in Australia seem stupidly high to people used to American car prices...)
Why the downvotes? I'm serious. If cars start lasting 600k+ miles and are shared by multiple drivers and families, car companies will ‘need’ to recoup their losses. Expect prices of car components (the stuff that wears out) to skyrocket, and patent litigation to get rid of the after-market/3rd-party compatible components.
I don't think you even have to imagine any malice or profit-protection on the part of the manufacturer to see that.
The simple fact that current cheaper parts are going to be replaced with higher-quality parts is going to cause a corresponding rise in maintenance and replacement costs.
Maybe. Keep n mind these's a lot of parts of a differential (or motor) that _dont_ need replacing when reconditioning. In a diff, there's maybe 9 bearings, the pinion and crownwheel, and maybe the 4 spur gears. If you start with an undamaged but worn out diff, replacing those parts effectively gives you a brand new diff. The bearings are standard industrial parts worth maybe $40 or $50 ( at retail prices) and the auto manufacturers can't affect the cost/margin on them. The crownwheels, pinions, and spur gears are already all available from aftermarket manufacturers for any model likely to be used as a cab (at least here in Australia).
cars wear out from two things
1) age
2) use
A car, left sitting in the driveway for 30 years, unused and unmaintained, is unlikely to work very well or for very long. Rubber components like hoses, wire insulation, weather stripping etc become brittle and break.
As some other posters mention, cabs with 650,000 miles are not unheard of. I had a Toyota Landcruiser with 450,000km on the second engine, over 900,000km on the body.
A car that got 20x use would not wear out 20x as fast because a large part of a car wearing out is just age, not miles.
I recently saw something pointing out another wrinkle about taxis -- per mile, their engines go between hot and cold a whole lot less, so there's less stress from thermal expansion. Once they're on, they pretty much stay on for at least the driver's normal workday.
I don't know how much difference this actually makes, but common knowledge amongst the people I know seems to be that the most stressful time for a normal engine is starting up.
It's not just common knowledge, it's the truth. And it makes a huge difference.
When an engine starts up, all the oil is sitting in the bottom. In a good condition engine, the oil pump starts giving meaningful pressure the moment the starter turns, and starts pumping fresh oil throughout the engine.
However, when cold, an engine has the wrong tolerances to account for when the materials heat up and the materials expand. So the oil pressure isn't quite right.
As the engine ages this problem gets worse, so each startup cycle gets progressively worse. This is why a car with a worn engine will show the 'oil pressure' light for an increasingly long time after it's started.
The length of service life for an engine will come down to
a) operating hours (not just distance)
b) operator abuse (revving while cold, excessive RPMs throughout use)
c) service attention (oil changes, filter changes, coolant changes)
d) duty cycles (how many times it heats up and down).
The worst thing you can do for a car is a lot of short trips with a big enough spacing to let the engine cool, and aggressive driving while the engine is still cold.
An F1 engine is seized when cold, it requires several hours of warm water and oil to be pumped around to bring up the metals to the operating temperature.
Modern engines can go a very long way if cared for properly.
Whether urban legend or not, supposedly there's a mercedes "million miles" club. They did famously buy back a mercedes that had over 2 million miles on the clock that's in their museum now
There's a Volkswagen Gol here in Uruguay with a million kilometers, and it's not an unheard-of amount
Cars are the most expensive in the world here, so we tend to keep then way long past their expiration date - as an example, I own a 1994 Maruti with 200.000 km, they aren't designed to last that long ! Japanese cars are the most coveted because they do last a million kilometers if cared for properly.
Sadly, there's a ban on used car imports (The vice-president's campaign was funded by the new cars importer association).
A million kilometers is considerably less than a million miles. A million kilometers is about 621 thousand miles. That's a lot, but it's much less than a million miles.
With my comment, I just wanted to add one more anecdotal point :)
I'm aware of the difference between miles and km (though I instinctively tend to minimize it and believe the difference is less than it really is)
I know of that Volkswagen because a million km is headline-grabbing here (on the "anecdotes" section), there are probably cars with a million miles but 1,609,344 km is not a headline-significant number, much like 621,371 miles isn't for the US.
I did actually find this quite helpful, as a Canadian I usually gloss over the km/mile conversion with a nice 1 km ~= 1 mile, as with most things you talk about (speed < 60, distances less than 100) the difference is fairly miniscule (and handwavey!). I didn't realize 1million miles is only 621,000 km! Thanks from an ignorant Canadian :)
As another Canadian, it's annoying to have you link your ignorance to your nationality. I'm confident that the vast majority of Canadians know that a mile is, very roughly, around 50% longer than a kilometer.
And if you think the difference is miniscule in normal usage, you just try driving 59 MILES per hour on a city street with a 50km/h limit, and see how the cops feel about that. (I say 59 because my lived experience is that everyone drives 10 over, and the cops don't ticket at 9 over; your city may vary)
Or try estimating when you're going to arrive at a meeting that's 100 MILES away on the highway, when you think "oh, 100 km, I can go 110 on the highway, plus the time to get out the door, call it an hour".
Even at walking speeds, 2 miles of walking is going to feel different than 2 km of walking.
Rust qualifies somewhat under #2, but not completely. Rust isn't 100% preventable in some areas. I live in Michigan and because we use salt on our roads, if you're not washing the vehicle (along with the undercarriage in particular), you will inevitably get rust. On my last vehicle the engine mounts (which were part of the frame -- unibody) rusted. There was pretty much no way to fix the car soundly. If I had washed the underbody frequently enough I could have saved it, but there is a cost associated with that as well.
I think it's safe to assume that cars driven 20 times as much wear out somewhere faster then the normal rate, but less then 20 times as fast. Some of the wear is due to time, some due to thermal cycling (turning on and off, which happens a lot less for the high-use vehicle) and only some is linear with use.
I was thinking along these lines on a similar thread recently. The particular thread implied that the cost would be much lower, because there was no driver involved. I agree that it would be lower, but not orders of magnitude lower, because the demand shape will be exactly the same - or even more pronounced.
My hypothesis is that the driver is probably about 10-20% of the cost of a fare, the rest is the capital cost of the vehicle + licensing fees + insurance, and the marginal cost of maintenance and fuel. Because inevitably cars sit around most of the time, then the price of 5-6 busy hours of the day has to make up for the rest of the time.
Further, with a disruptive business idea like this, I could easily see an auction-style interface for the vehicle booking, which would give a much better revenue curve (we are talking about Google). In that case, the peak-demand period would probably exceed the current (regulated) taxi fares. But the plus side of that is that a midnight ride would be very cheap due to lack of demand and simultaneous lack of a need to pay drivers more money to work nightshifts.
Yes, cars may still sit around doing nothing at night, but the number of cars for peak traffic will drop significantly because of the gains in utilization. Imagine you want to ride share with your friends, one car will ferry people to and fro the highway which will be almost impeccably timed with your friends arrival at a waiting area just off the freeway. You'll get in the car and continue to work, after you arrive at work that car will go pickup someone in the city who works a bit later and grab a couple of their friends on the way.
When you combine this technology with social networks and mobile the improvements to efficiency, cost, and quality of life will be astounding. This isn't going to be an overnight thing but I see this kind of thing becoming prevalent probably 5 to 10 years after the first driverless cars become publicly available.
Driverless cars would be worth it simply for the reduction in drunk driving.
Driverless cars would be awesome simply for the increase in drunk driving. This is a really good point actually, you could go out for the night and still drive home.
Drink driving laws have had a huge negative impact on country living in rural Ireland, where the main social outlet is the local pub, and there isn't a taxi in the village.
> if there are enough cars to handle peak traffic during the day
I got the image of cars migrating across the Eurasia, taking Chinese to work, then Indians, then Middle-easterns, etc.
If cars are getting 20 times more use they will wear out much more quickly than they do now.
This is just a matter of the practical design choices. Any machinery that is in heavy use usually gets fitted, appropriately, with more robust set of parts which last many times longer before wearing out.
Current consumer cars have bearings, joints, and moving parts that are carefully optimized to match the expected usage pattern (which is mostly idle) for a designated period of time and nothing more. That's why older cars can sometimes run for ages. Decades ago we didn't know how to make extremely light-weight parts from least amount of steel with a calculable expiry time of, for example, 40 thousand miles so engineers had to fit cars with slightly heavier and more expensive parts to make sure they didn't break too easily. Think about fitting bearings and joints from a heavy van into a light Japanese small car. Or consider old 70's-80's Saabs and Volvos that can last nearly forever.
Likewise for fluids and lubrication, it's easier to clock high mileages with a car that is mostly in use throughout the day rather than with one that is used a couple of times a day for commuting. The engine wearout is at its peak during the first miles after a cold start.
Also, the 1970s and 1980s were a low point for cars in general -- lots of new environmental regulations were kicking in (including a whale oil issue which was discussed on hacker news), causing objectively worse reliability for many cars in the 1970s and 1980s vs. the 1950s and 1960s. By the late 1990s the Japanese and German manufacturers seem to have resolved things, and by the 1990s, the US automakers.
340 comments
[ 3.0 ms ] story [ 290 ms ] threadOf course, with driverless cars, people may begin to stagger the starting / ending hours of the work day to allow for owning fewer cars.
If entity that provides those cars is big enough, they can match up the destination and optimize the traffic so that people going to the (roughly) same place share a car. When I was commuting from SF to MV for work, I had to use the whole car. There were plenty of other people who were commuting at the same time from/to very similar destination and used the whole car. It's not hard to algorithmically put several such people into the same car.
Today cars are 4/5 seaters so you get at 3x reduction compared to current levels (taking into account that not everyone drives by himself today) but you could easily redesign the cars to not be much larger and taking 8-10 people. Or make them straight up buses. Google already does that with their shuttles where they pick e.g. people from SF and drive them in big buses to MV, except it would be much more efficient because the potential pool of people transfered would not be just "people who work at Google and live in SF" but "every person who lives in SF and works somewhere in MV", which is a much bigger number.
Isn't this the travelling salesman problem?
http://en.wikipedia.org/wiki/Travelling_salesman_problem
Of course, a non-optimal, but good enough routing is probably fine.
EDIT: Added last sentence.
In any case, the problem more likely to be used in delivery is the VRP or a variety: http://en.wikipedia.org/wiki/Vehicle_routing_problem.
Of course, I don't have much firsthand knowledge of computational complexity myself, so someone else who know better should call me on any nonsense above. =D
What you refer to is the fact that the computational complexity (i.e. time to finish it) of known algorithms rises exponentially with the size of the problem and exponentially is a code word for "really, really quickly". It just takes too much time to find the minimal path if your graph is big.
What I meant, however, is that the problem has been solved in practice. When you ask Google how to drive from SF to NY, it'll give pretty good answer in milliseconds. Is it an optimal answer? It might be, it might be not, but it's a very good answer. Getting slightly better answer is not worth the computational time because it won't make a difference in practice in your trip.
Similarly, a car rental company doesn't have to schedule things optimally, they just have to schedule things really good, and that's possible with much less computationally expensive algorithms. The big win is when you go from "no optimization" to "good optimization", not from "good optimization" to "perfect optimization".
In retrospect, I probably should have kept my thoughts to myself, and I probably deserve the downvote or two that I got for not doing so. =)
That's not the traveling salesman problem, it's the shortest path problem, for which polynomial-time algorithms are well-known and taught to first-year computer-science students. (To be fair, driving directions do require coming up with good edge weights on the graph composed of the American highway system, but once you've got weights, you can run Dijkstra's algorithm and you're done. That's probably not how GMaps driving directions work, but the point is that driving directions are not TSP.)
For example: 1. Drive to work 2. Drive home 3. Drive to school 4. Drive home 5. Drive to store ...
Many parents already drop their kids off @ school and thus are adjusting their schedules appropriately.
It'd require inter-household car sharing on a significant scale to take advantage of that, though.
And even during peak times, car rentals will be augmented by a suddenly-reinvigorate public transport system; your car may take you to a bus, to be picked up by a car on the other end.
The only missing bit of tech is the actual self-driving car. Everything else is simple and obvious extrapolation from brain-dead simple Economics 101 and already-existing network technology. Modern car usage will be seen as only an incremental improvement on horse-drawn buggies, with the real revolution happening much later.
I don't know the exact breakdown but my guess is that salaries of drivers are much more expensive than buses and gas, especially given those are often unionized jobs with very costly benefits (pension etc.).
I believe the main reason local government provides public transportation is because to achieve low prices it needs to be subsidized. No private enterprise is interested in competing in a market where you can't make profit.
At the same time, you don't see local government e.g. opening grocery shops: this need is well filled by private enterprise.
The car rental business would develop despite public transportation because it covers a much bigger needs. However, when it reaches scale and the low price that comes with it, it might make public transportation no longer needed as it'll also cover that use case.
Even if it's slightly more expensive, would you rather pay $2, wait 15mins, not get exactly where you want to go and hope no one will punch you in your face and steal your iphone (true story!) or pay $5, wait 5min and get exactly where you wanted to go?
If enough people start doing that, at some point the fixed cost of running public transportation will be too high to justify low usage by public.
Not to mention I do expect for profit enterprise to adopt driverless cars much faster than government agency, hence giving them a cost advantage for quite a few years.
If enough statistical evidence accumulates to demonstrate that these are safer than manually controlled cars, look for the insurance companies to impose a severe cash penalty on driving the car yourself. If the difference is large enough, you might find yourself unable to purchase insurance for a manually controlled vehicle at any price.
In a few scenes in the movie, the female lead (the robot psychologist?) reacts in horror to Will Smith manually driving his petrol driven motobike because it has not auto-pilot and 'you realise petrol is explosive don't you?!?'
And after Will Smith has the forced accident in the tunnel (when attacked by the robots), his boss accuses him of unsafe driving because the blackbox reported that he had disengaged the auto-pilot.
So quite conceivably we'll end up in a future where manual car driving is seen in the same light as drink driving or speeding is seen today.
http://en.wikipedia.org/wiki/Jevons_paradox
The key to the Jevons paradox is that first order estimates of the effects of efficiency are misleading. You must also give regard to how the increased efficiency will change buyer preferences.
You'd find automated commuters living way out of town in areas that are currently nearly uninhabited. There probably aren't many places in the lower 48 that are too far from a medium-to-large city to commute with automated driving.
Road trips would also be easier - you'd have to be in a pretty big hurry to pay for an airline ticket when you can go 1500 miles or so in 24 hours with no effort, and without the need to pay for a hotel.
That probably wouldn't result in a large net increase in total fuel consumption.
According to wikipedia the average commercial jet gets 49 passenger miles per gallon. It wouldn't be to hard to get that out of a driverless vehicle, add a second passenger and you're twice as fuel efficient as an airplane.
If you did want to pay for a car to swing by your house to pick you up, it probably wouldn't be a car, but something more like a small bus with sleeping compartments that could pick up and drop off people along the way.
If you could fit 10 people along most of the route (and you could assuming it was between 2 moderately populated areas and road trips were as common as they probably would be), your passenger miles per gallon would be incredibly high.
Combine driverless cars with maglev trains that can combine multiple cars into one long train with individual cars coming and going, and you may have something. I'm not holding my breath that I'll see anything like that in my lifetime. Transportation infrastructure in the US has gone nowhere in my nearly 40 years; I don't expect a great deal of change (to the infrastructure) in the next.
I'll calculate the costs 10 passenger van, if you could work it out so that it gets an average of 40 mph for say 20 hours out of the day. That's about 300,000 miles per year divided by 20 miles per gallon that's 15,000 gallons of fuel. At $4 dollars per gallon thats $60,000 per year in fuel costs.
I'll assume another $60,000 per year in maintenance and depreciation (this is probably high b/c you can buy a new large passenger van for under $30,000)
We're at $110,000 in variable costs so far.
Now at $30,000 per year per driver in total costs (could be less, could be more depending on area)
Driver works 40 hours per week with sick time and vacation time. You'll need around 5 drivers in order to keep the van operational 24 hours per day.
That's $150,000--more than half of the total costs.
There are also other factors to consider. Drivers, don't want to operate that far from home, so you have to pay more.
You have to set up a logistical solution so that you can refresh the driver every 6-8 hours. That means depots where you can exchange them.
I'd say there would be significant cost savings.
It's still rare for people to commute that far on a daily basis. Everyone I know who travels more than about 2 hours from home to work tends to take weekday lodgings within walking distance of work.
A 3 hour commute will always be a big deal, because even if you work a strict 8 hour day, never staying late, that means that 14 hours out of 24 are spent away from home. If 8 of those remaining 10 are spent sleeping, you now have 2 hours left in which to indulge in any kind of leisure or childcare activity. Not everything can be done on the move - I can't imagine it being very likely that you'll be able to go for a swim or play rugby or cricket whilst travelling to and from work.
Also there's no reason why we need to keep the concept of a 4-5 passenger sedan. 10 passenger vans could be viable.
>free automated delivery
could be routed so that it's delivered to you by the car that's coming to pick you up in the morning to take you to work.
There's also no reason non-passenger vehicles couldn't be ultra efficient 35mph crawlers.
By taking the driver out of the equation we can drastically our definition of a car.
Potentially a lot. In a 100% automated environment you don't need stop lights and traffic jams are substantially reduced. But yeah, I agree that the increased use resulting from greater convenience would outweigh that.
Also, why would you want to restrict an improvement that could benefit gas powered cars too? A self-driving car would probably be more efficient than a human driver. Fewer traffic jams would be less wasted fuel.
It would be good to get the data about energy consumption for car building, to answer this.
In that scenario, both the cost of cars and cost of energy (gas or electric) will have direct negative impact on their revenues so they will have a very strong economic incentive to drive down that cost.
A single buyer isn't well informed and isn't very rational, so a nice shape of the car might be more important than mpg mileage. Our hypothetical provider would, however, consider it one of the most important characteristics.
Combine that with a great purchasing power of such provider. If they buy 10000 cars and they want high mpg, you can be sure that car companies will spend most of their effort on increasing mpg.
I want one and I have money - when they're available, why wouldn't I be able to buy one?
Are you suggesting that I won't want to? I'm skeptical - the convenience of being able to use something whenever I want trumps cost. Or are you suggesting that I won't be able to?
Once it's a proven small-scale model, larger transit and freight systems will start using them; I'd expect ZipCar and Uber and similar services to jump on it. At this time, more and more people would be buying their own driverless car.
Now picture 20 years down the road. You're a young person who can't afford a car. Do you get a job to pay to buy a car, or do you use a service like ZipCar? Fast forward 5 years, you've got a job and you could afford a car. Do you buy one to give you "added convenience"? That's what ZipCar will have to beat. If they can get a car to you by the time you're outside your house, they are actually beating personally owned vehicles, and the closer to that they get, the less likely it is people will purchase vehicles.
- comfort while riding (lack of abrupt motions requiring me to brace myself)
- courtesy to other drivers and pedestrians
- actual number of traffic accidents (colleagues: 0, taxis: 2, although fortunately both accidents were mainly property damage rather than serious injury... but still)
I think you would, but the conventional wisdom around owning your own car would along the lines of the current CW surrounding owning your own aircraft: it would be something for wealthy enthusiasts, not the general public.
Some very rich people still own private jets, but it's rarer than it would be if more people were rich.
1. You have to change oil, tires and make all the other repairs.
2. You have to pay for insurance.
3. You have to pay for parking. In SF parking alone can cost you $100/month (I mean if you don't own a house with a garage (which will take you back $1 million); I don't think most rental apartments come with parking spaces so you either have to stress out about finding a parking space for the night close to where you live or pay for parking spot).
4. Parking while driving in the city. If you need to drive somewhere and stay there, you have to find parking, which, in SF, is not easy to do. A rented car just drives you to where you need to go and then goes of to pick up the next customer. When you want to get back, you summon another car.
5. What if you want go skiing for the weekend and need snow tires? Would you rather put snow tires yourself or summon a car that has them?
I'm in SF and I already use my car so rarely that its battery died on me. I do ZipCar but it doesn't cover a lot of scenarios that, while not frequent, do happen and the requirement that you need to return the car to where you took it from makes it non starter for short trips within city.
If there was a service where I could just summon a car anytime, it would be cheap enough to work as a taxi replacement but also for longer trips, I wouldn't buy a car.
I'd very happily own a driverless car. Even better will be when I take it out of driverless mode in the rural roads. It would be able to take itself to get its oil/ires serviced,to be recharged and to park somewhere cheap and out of the way. Bring it on.
See: Pocket watches > wrist watches > cell phones.
I do that.
But it's not because I hate planes but I've got three kids. Airfare for five to visit grandma in Virginia is prohibitive compared to the cost of driving.
I'm not a nut about it: if it's just my wife and I, we fly.
But on those long drives especially, I'd love a car to do the driving.
[1] http://dsc.discovery.com/cars-bikes/do-driverless-cars-offer...
It's quite reasonable to expect that it will be economical to pay something like 20% more for the drivetrain to make it more durable, but that's not a radical re-design.
But as a complete solution - which is what the article talks about - the cost of 'wasted' fuel is negligible given the change in behaviours and travel methods that automated cars may bring about.
Cars wear out primarily through driving. If you simply switched to automatic driving, what you would have is fewer cars being driven more often and so being worn more quickly. If car-mile consumption stays the same, new-car production would stay the same. On the other hand, if the auto-drive cars increased carpooling, then you'd see a decrease in car-miles consumed and so a decrease in production. But if auto-drive cars drove around empty more, you might have even more car-miles being consumed.
Moreover, you'd have a "big bang" where people decide to mostly stop driving the old, non-automatically-driving cars and so there'd a huge spike in consumption at that point.
The space saved by avoiding parking could be really large, still.
A nice thing would be that at the start, a person might be able to finance their self-driving car by renting in out when they didn't need it. Those economies might make the phenomena spread really quickly.
At the very least, I expect that would be offset by some amount by the fact that automatic cars are going to be driven in a much more efficient and car-friendly manner than they are when a human is operating it with their imperfect motor skills, their ego, their laziness, etc. The way the cars operate would be (or at least could be) algorithmically optimized so that all those little things you're encouraged to do (like pretend there is a glass of water on the dash when you're driving, etc.), actually happen.
Such performance is not free. It causes engineering tradeoffs that increase maintenance costs, complexity and of course reduced fuel efficiency.
But it's easier to sell a driver car with some oomph than not.
There will be other design tradeoffs you and I have not thought of (and could not think of) that will add up to further efficiencies.
My assumption here is that there would emerge a big car provider who would provide those cars, similar to how there's only few major car rentals, or how ZipCar is leading in its category or even how Uber emerges as a single taxi-like provider. The economies of scale are very big here.
Given that, some things you have not accounted for:
Higher average utilization per car. I don't know how much commute traffic accounts for total traffic, but by judging rush traffic on 580, it's quite a lot. Most of those people go from the same location (e.g. SF) to the same location (e.g. Mountain View) and today most of them drive alone. Today 4 or 5 of them would be scheduled by an algorithm to use the same car.
Because of that the cars would get re-designed for a higher capacity (at least 8-10 people) or even straight up buses. A big provider would have so much data that they could optimize the hell out of sharing cars.
Similarly, the cars would utilized, on average, much better. Again, a company with enough data could really optimize for a lifetime of a car and they would have economic incentives to do so (unlike a single customer for whom even gathering the necessary data would be cost prohibitive).
Our cars are, without a question, utilized very inefficiently. Someone operating a fleet of tens of thousands of cars would have not only economic incentive but also necessary data and necessary transaction volume to optimize per-car utilization, because for them it would translate directly into large savings.
We no longer need to worry about designing it so that the driver can see every angle, so we can design more efficient seating layouts. We can even add partitions so the 10 passengers don't have to see each other.
Long distance deliver vehicles can operate at lower fuel efficient speeds.
We can combine delivery and passenger vehicles.
The list is endless.
To your other point: I think that assuming we'll go off the wall and design wacky new vehicles is too optimistic, akin to the "cities of the future" envisioned during the past hundred years. People's sensibilities of "what makes a car look nice" and "what is a metal box I'd like to sit in" seem to be pretty set, and we don't have a very different experience in a train or on a bus today than in a car. Face forward or backward in a seat with big windows and high visibility.
Part of that is for your senses to not freak out, part of it is just that it's cheap and easy, and part of it is that experiments in the form-factor don't need to be done, so they aren't done. I expect the same trend will continue unless flukes occur.
I didn't know that about greyhound. I'm aware that airliners routinely carry packages. I was thinking more, you've got a car coming to your house every morning, let's throw your Amazon.com orders on it.
>design wacky new vehicles
I'm not thinking wacky designs. A car is usually designed to be sat in for a certain length of time. They aren't really designed for very long trips without stopping frequently, since driverless cars make long trips much more appealing, they'll have to be redesigned to an extent.
A fleet of rental cars would be much more efficient if each car held say 10 people instead of 4, so another natural redesign.
> "what makes a car look nice"
To many people, a car becomes part of their identity, not so much with a temporary rental, so I think function starts to trump form
>don't have a very different experience in a train or on a bus today than in a car
I think the experience in a long distance train is very different than in a car.
The main difference is compartments, and that's what I envision for cars. I'm not talking anything radical, just what I think would be natural progressions when you remove the need for a driver.
Software would decide routes and timing to best service as many people as possible.
While I'm all for your point that drivelessness will allow an incredible re-imagining of the automobile, considering the gp responded my original post, I just want to bring things back to the question of resource utilization.
The automobile today is half efficient transport and half personal expression/personal entertainment. And the personal expression/personal entertainment part is where the massive resource utilization comes in. So, sure, you could reimagine the automobile for super-efficiency with four-people per car whenever you're driving and the resource utilization goes away - so does the personal entertainment/personal expression stuff. So you could go multiple ways. Towards a super-efficient taxi and towards an office/living room on wheels. The first way would involve less resources consumed, the second would involve less resources consumed. It is hard to be certain what the net outcome will be.
To many people, a car becomes part of their identity, or like you said
>expression/personal entertainment
I'm sure that there will still be people for whom this is true, but for most of us it won't matter b/c the efficient always there taxi will be so much cheaper.
For most middle to lower-middle class Americans I think current cars are really out of their comfortable price range, they own them b/c they view them as a requirement.
If you give them an alternative that is just as convient, but cheaper, and without the maintenance hassles, it's no contest.
Taxis are for in-city travel and occasionally nearby inter-city travel, and the regular car has ended up being fine. They do have a few larger vehicles for big groups, but those are special case vehicles. On average, they tend to get 1-2 people at a time wanting to go places.
Airport taxis tend to be closer to a commute-oriented vehicle. Several people are going from here to somewhere else, so they tend toward 5-6 person vehicles which operate through the night on >1hour long trips.
Corporate transit systems like Microsoft's Connect system are essentially multi-user multi-destination. Those make buses make sense, but there are a relatively small number of buses making set route transfers while a hive of taxis provides last-mile service.
The "Future of Transportation" article from a few days ago [1] suggests that the average vehicle has 1.2 people in it, and that apparently takes commuters into account. Now we must consider: will behaviour change to increase carpooling?
Specifically, what is presently preventing people more people from taking buses to work? (lots do take buses and trains already, but many people drive 10 to 100 minutes)
It could be that a bus doesn't get you /right/ to work. It could be that you would never think to contact everyone in your neighbourhood and figure out that you could buy a bus and run it just for you guys. It could be that the cost of owning and using a car is just low enough that it doesn't really save you anything. It could be an aesthetic sense of independence.
My bet is that this will work out like Microsoft's system. Major transit routes can scale passenger levels up and move efficiently between well placed bus-station/pickup hubs, and armies of taxis will shuttle you to your final destination. Google's bikes are nice, but last-mile service will have to deal with a broader sprawl. Most of them will be focused on getting from hub to delivery and vice versa during morning and evening, but then, what happens during the day? Even weirder: what happens at night?
1. http://swiftprt.com/blog/2011/12/the-future-of-ground-based-...
edit: Also, to build on the above comment, the benefit of having a car is having the freedom to use it when you want, it's always there. Driverless cars always in motion are basically public transit. Just take the bus/taxi/subway already.
Minus the cost of the driver.
The point of ubiquitous driverless cars, is that one could get to you so quickly that for all intents and purposes it would always be there.
Also they don't always have to be in motion. Just give people a discount on service if they let cars park in their driveways while waiting for instructions.
http://en.wikipedia.org/wiki/Luddite
Different jobs will take their place. There are a lot fewer farmers these days, for example. The real trick is to find effective ways for people to get the training.
This is the wrong place to make this argument (it should probably be made at length) but I think the only real solution when we have a super-efficient, highly automated world is to drastically increase taxes to the super-rich and subsidize people with obsolete skillsets. But regardless, we'll no doubt see more interesting developments here in the next ten years. The current unemployment situation is just the beginning, although a lot is going to happen in the meantime.
(This is from a country where about 10% of the population is currently on welfare and we almost de-facto already have the economic subsity system I'm describing)
http://motherjones.com/kevin-drum/2011/09/who-will-tend-mach...
http://arstechnica.com/old/content/2008/10/future-of-driving...
As the efficiency with which a resource is used increases, the tendency is to use more of it rather than less.
Concretely, if I don't have to drive the car myself, I wouldn't hesitate to drive anywhere. I'd go to the city every day if I could read on the way, and sleep on the way home.
Today, car owners have already spent $x on their cars -- so they have an incentive to use their investment to the fullest.
Compare that with a shared-car system like ZipCar. ZipCar is generally much more efficient than owning a vehicle (if you don't use it to commute) -- however, since you pay per hour, you have an incentive to skip the car when it's not necessary.
By removing the sunk-costs of ownership, you could incentivize conservation rather than consumption.
And if even 10% of drivers switch to using this model, then there's always going to be a car on your block. That's pretty competitive.
With driverless cars it's even less of an issue, because the car will literally drive from the end of the block (where it's waiting) to your door.
Assuming, that the San Diego prices can be applied to Francisco, it seems like taking such a car from downtown to the airport would be about the same as the BART subway. I assume, that would be too expensive for a daily commute.
But with a self-driving car, you'd skip both steps. Instead, you'd ask for the car and step out in your driveway when you get an alert that it's here.
There is surely an upper bound to how much traveling any person wants to do. If it were 5% more convenient for me to travel by car, I _might_ travel 5% more, but I couldn't see myself traveling 50% or 100% more no matter _how_ convenient you could make that travel.
Lets give it 15 years. 15 years is a long time for technology but not so much for societal evolution.
I can see the driverless car being on the road in 5 years and I see it being used for carpools. But the vision as painted in the post? Yes, I'm a skeptic.
Very interesting stuff though...I can't wait till they hit the road.
Who is liable if your driverless Civic drives you into a tree? What if it runs over a pedestrian? Honda? Google? You?
Driverless cars could dramatically reduce the dangers of driving, but it's a lot more complicated if a machine learning algorithm kills somebody than if a mistaken human being does.
All it needs is for someone to take liability. If you own the driverless car, then you buy the insurance. If it's leased from a pool, then they buy the insurance.
The software vendor for their part take an over-arching insurance on software errors across thousands of cars. Done deal.
Today if you have a 20 year old your premium is higher. Tomorrow if you have a driverless car, your premium might be low even if you have a 16 yr old. Parents buying a first car might just buy a driverless car for a lower insurance premium & the safety of not having the kid drive!
http://mattmaroon.com/2011/01/03/google-will-become-an-ai-co...
But I think the removable trunk will be combined with a passenger-less unit that carries the trunk back to your house and drops it off. Sort of a tiny version of container shipping.
Lots of people do already use mass transit every day, which has a cost structure closer to driverless taxis (because the labor cost of a bus or train driver is amortized across more passengers).
Incorrect conclusion. Cars being used 20 times more will break down 20 times faster. Less might be needed on the road at the same time but they'll break down and need to be replaced at a much faster rate. Over a serval year timeframe the amount of cars turned over will remain roughly the same.
I also wouldn't send my car home to pick up my wife and drive her to work because my fuel bills will skyrocket through all the extra driving around.
The shared resource makes sense. It'll be similar to taxi services that exist now. You call for a driver and the closest shared car will come to pick you up.
Not true if they're owned by big players who professionally maintain them. Want to buy a used LLV (postal truck from the eighties/nineties)? You can't, because they haven't been surplused yet.
Also, cold engines wear much faster than hot engines, so an engine that runs all day does not wear out as quickly per mile as one that's driven a few times daily.
You make a good point that a fleets maintained better than the average citizen but they will still break down and still need to be replaced faster that the author is stating. We will still have roughly the same amount of turnover over the course of a decade.
The reason? The economics of electric cars and parking.
Initially it seems that they will fix the parking problem - driverless cars can be sent to park outside the CBD, reducing traffic and freeing up space.
But the problem is that people/software will optimize for price, and for electric cars the cheapest scenario is for them to be stuck in a traffic jam on a public road.
Instead of going to a parking bay, the software will route them to the nearest traffic jam, where the car can sit with the electric motors off for a large amount of time. Inevitably, some software will misjudge how long their charge will last, their batteries will run flat and the traffic jam will get worse.
As far as I can see the only way around this is to increase the cost of being on the road.
Even if groups of hackers installed this on their cars, I can't imagine this movement gaining enough critical mass to honestly be a concern.
What's the risk, exactly?
Our car has an "ultra-economy mode" where the motor turns off when stationary - many cars already have this.
Our car can park itself - many cars already have this.
Our car will find a parking space, optimizing the route to and from the parking garage - that seems logical, why would a company not do that?
Our company will optimize the route to and from the parking garage, and the time spent in the garage vs on the road based on price vs economy of the car - why would a company not do that?
The problems only occurs when hundreds of cars independently make their own decision to drive as slowly as possible to and from the parking space.
(Edit: someone else pointed out that UPS routes their trucks to avoid left turns. This isn't too dissimilar to that - they do it to improve fuel economy, this would be done to improve costs too)
Bad press for all involved.
If you made it a flat fee, then it's kind of like runs-batted-in. The car has to get more people delivered to make money, so once it gets you to where you need to go as efficiently and quickly as possible, you get out and it speeds off to complete another fare.
Increasing road-cost would have the opposite effect, I think. If the cars make money just by being on the road and not by completing tasks, they lose the incentive to finish and jam-sitting becomes optimal.
But I'm not optimistic about the share-car thing, at least in cities without a strong existing public transport system. In most cities too many people travel at rush-hour, with unique route requirements (dropping kids at school etc).
Driverless cars will still be limited by physics (their size and how fast they can accelerate). If car routing algorithms deliberately route cars to places where the traffic is slow (because they can loiter, saving on parking while using little power) then the traffic jams will be worse.
If we rebuilt the road system specifically for driverless cars maybe we could eliminate traffic jams. Otherwise...
Secondly, it also opens up the scope for private taxies. A person can allow his car to be utilized by a certain section of his peers - colleagues, neighbours etc for low prices - resulting in lower parking time.
Yes, and for the forseeable future there will be conventional cars, as well as bicycles and pedestrians.
Roughly, 50% of my fuel bill is on direct taxation. That's it's primary tax cost, not the fixed annual license cost, and gives me a tax bill that's elastic by usage in exactly that way while also promoting efficiency.
Of course, best wishes to anyone trying persuade the average man in the street that this is A Good Thing....
They can turn off their motors without losing any efficiency (which is why hybrids do so well on efficiency in city traffic). Depending on how slowly the traffic moves there will be a point where it is cheaper to stay in traffic (with the motor off x% of the time) compared to paying for parking.
Cabs in the West are expensive because drivers need to be paid more and cartels are at work. If both these costs are eliminated then what works in a country like India is, suddenly, the best option for commuters in the West. Realtime pooling of cab passengers is a possibility, a traveler could be offered a choice of picking a co-traveler for a reduction in fare(or even cash back!) when already travelling in the car.
Families could still own cars, but they will probably pay for insurance on a per-mile basis, rather than a per-month basis, so that the incremental cost of owning a car is minimal. Effectively, you end up owning a cab that only you use and pay cab fare to the insurance company :-) The ratio of cabs to private cars would then be an interplay of insurance costs/parking costs and ride sharing benefits, with cities leaning towards cabs to escape parking costs. So, things wouldn't be too different :-)
One unfortunate consequence could be the increase in suburban sprawl and traffic congestion with driverless cars as people start caring lesser about longer commute times. Automated cars will probably be better behaved in traffic, but the road capacity will be pushed to its limits.
Most traffic in the US is caused by the lack of roundabouts.
In any event, it's as likely as seeing people riding segways all over the place.
Just walk outside your house and a car drives on up, opens its doors, and takes you to the most convenient location to transfer to a public bus, or if economical to pick up other passengers going near your destination.
And if cars are receiving 20 times more actual use, that would imply that there would be 20 times less cars sold
Actually, no it wouldn't. If cars are getting 20 times more use they will wear out much more quickly than they do now. That means cars will have to be replaced much more frequently. There would be fewer cars sold than there are now, but it wouldn't be 20x fewer.
The operating percent of a car will go from 4% to that 96%
This seems wildly optimistic to me. The driverless cars may be capable of driving around 96% of the time, but that doesn't mean they can be carrying people 96% of the time. No matter how efficient the system, if there are enough cars to handle peak traffic during the day, then a lot of those cars will be sitting around doing nothing at night.
If my apartment complex had a few dozen cars shared between a few hundred apartments, perhaps the car I take to go shopping could pick up the next door kid from soccer on the way back, then a different car might pick me up at the shops when I'm done, after dropping some other neighbour at the movies...
(I guess I'm now describing taxis. I wonder what the difference between this, and a taxi network of driverless cars is?)
Anybody got both Travis Kalanick and Elon Musk's numbers in their speed dial?
What if a "disrupt the cab industry" company got together with a "low moving part, high reliability electric car maker" to do an end-run around the expected auto industry opposition...
A fleet of driverless electric taxis, all routed by smartphone apps and behavioural prediction...
During the middle of the day and all night, you lease the capacity to FedEx or UPS...
Maybe it could be coupled with swappable battery pack stations that the autotaxis could visit to get a freshly charged pack. A geographically focused taxi company could have the financial resources to invest in the battery depot, which would also solve the standardization problem.
If we stop owning cars and the communal ones we use drive themselves home, we'll stop putting money in parking meters.
Replace the parking meters with charging stations...
http://www.ted.com/talks/shai_agassi_on_electric_cars.html
http://www.globes.co.il/serveen/globes/docview.asp?did=10007...
Lower utilization is balanced by lower energy costs than gasoline (but you're right, swappable batteries would help too).
http://www.brightautomotive.com/
You can't have a car with a loaner carseats either since they have to be individually adjusted to the kid.
Every single time people talk about cars, and public transportation they always forget about kids. I see it over and over. Come on people - expand your worldview a little.
Why not have a percentage of driverless cars pre-equipped with carseats? You'd just have to configure once the age of your kids, then when you call a car (using e.g. a smartphone app), you just say for who it is ("Siri, I need a car for john, mary and me") and the right type will come.
And you're going to need every combination of ages, people have more than one child.
There are approximately 10 different car seat configurations for the various age ranges (5 actual car seats). And assuming up to 3 kids would require 1000 different cars.
There is rear facing (with 3 heights), front facing strapped (4 heights), front facing buckled, booster, and booster without back. (Although many car seats can handle 2 types in one seat. But it requires you to reinstall it.)
But converting the seat requires rethreading the straps and other adjustments - it can take an hour to install some types if you are not familiar with it.
Existing cars, if driven 20 times as much, would wear out 20 times more quickly, like taxicabs do now.
But it's also possible that cars would simply be built with more reliable components and more durable materials, like current aircraft and public transit vehicles are.
It's not cost-effective to build an ultra-reliable car that's sitting idle 96% of the time, but the economics would surely change if the utilization rate is much higher.
I remember talking to an old-school cab driver a while back, when I noticed his odo had ~650,000km (~400k miles) on it. We chatted a bit about it, and when I asked "So how long do cabs last" he said "3 good crashes." It doesn't which of the locally popular cab models you buy and it doesn't matter how far you drive them - you might need to fit a reconditioned diff or gearbox or even motor, but all of that is "routine maintenance" from his point of view. Its after the third time you've crunched it hard into something that it's time to get rid of it...
(Yeah, that's what the base model cars used for cabs cost here. Google Holden Commodore or Ford Falcon prices. New car prices in Australia seem stupidly high to people used to American car prices...)
The simple fact that current cheaper parts are going to be replaced with higher-quality parts is going to cause a corresponding rise in maintenance and replacement costs.
As some other posters mention, cabs with 650,000 miles are not unheard of. I had a Toyota Landcruiser with 450,000km on the second engine, over 900,000km on the body.
A car that got 20x use would not wear out 20x as fast because a large part of a car wearing out is just age, not miles.
I don't know how much difference this actually makes, but common knowledge amongst the people I know seems to be that the most stressful time for a normal engine is starting up.
When an engine starts up, all the oil is sitting in the bottom. In a good condition engine, the oil pump starts giving meaningful pressure the moment the starter turns, and starts pumping fresh oil throughout the engine.
However, when cold, an engine has the wrong tolerances to account for when the materials heat up and the materials expand. So the oil pressure isn't quite right.
As the engine ages this problem gets worse, so each startup cycle gets progressively worse. This is why a car with a worn engine will show the 'oil pressure' light for an increasingly long time after it's started.
The length of service life for an engine will come down to a) operating hours (not just distance) b) operator abuse (revving while cold, excessive RPMs throughout use) c) service attention (oil changes, filter changes, coolant changes) d) duty cycles (how many times it heats up and down).
The worst thing you can do for a car is a lot of short trips with a big enough spacing to let the engine cool, and aggressive driving while the engine is still cold.
An F1 engine is seized when cold, it requires several hours of warm water and oil to be pumped around to bring up the metals to the operating temperature.
Modern engines can go a very long way if cared for properly.
It's not isolated to Mercedes either
See http://en.wikipedia.org/wiki/Car_longevity
http://news.google.com/newspapers?nid=2199&dat=20021117&...
He took extremely good care of it.
Cars are the most expensive in the world here, so we tend to keep then way long past their expiration date - as an example, I own a 1994 Maruti with 200.000 km, they aren't designed to last that long ! Japanese cars are the most coveted because they do last a million kilometers if cared for properly.
Sadly, there's a ban on used car imports (The vice-president's campaign was funded by the new cars importer association).
I'm aware of the difference between miles and km (though I instinctively tend to minimize it and believe the difference is less than it really is)
I know of that Volkswagen because a million km is headline-grabbing here (on the "anecdotes" section), there are probably cars with a million miles but 1,609,344 km is not a headline-significant number, much like 621,371 miles isn't for the US.
I did actually find this quite helpful, as a Canadian I usually gloss over the km/mile conversion with a nice 1 km ~= 1 mile, as with most things you talk about (speed < 60, distances less than 100) the difference is fairly miniscule (and handwavey!). I didn't realize 1million miles is only 621,000 km! Thanks from an ignorant Canadian :)
And if you think the difference is miniscule in normal usage, you just try driving 59 MILES per hour on a city street with a 50km/h limit, and see how the cops feel about that. (I say 59 because my lived experience is that everyone drives 10 over, and the cops don't ticket at 9 over; your city may vary)
Or try estimating when you're going to arrive at a meeting that's 100 MILES away on the highway, when you think "oh, 100 km, I can go 110 on the highway, plus the time to get out the door, call it an hour".
Even at walking speeds, 2 miles of walking is going to feel different than 2 km of walking.
Edit: PS, who's Sunbeam?
Citation needed. I don't think they would.
My hypothesis is that the driver is probably about 10-20% of the cost of a fare, the rest is the capital cost of the vehicle + licensing fees + insurance, and the marginal cost of maintenance and fuel. Because inevitably cars sit around most of the time, then the price of 5-6 busy hours of the day has to make up for the rest of the time.
Further, with a disruptive business idea like this, I could easily see an auction-style interface for the vehicle booking, which would give a much better revenue curve (we are talking about Google). In that case, the peak-demand period would probably exceed the current (regulated) taxi fares. But the plus side of that is that a midnight ride would be very cheap due to lack of demand and simultaneous lack of a need to pay drivers more money to work nightshifts.
Except possibly around whatever local time the bars close.
One other benefit would be that (hopefully) cheaper late night rides would cut down on drink driving.
When you combine this technology with social networks and mobile the improvements to efficiency, cost, and quality of life will be astounding. This isn't going to be an overnight thing but I see this kind of thing becoming prevalent probably 5 to 10 years after the first driverless cars become publicly available.
Driverless cars would be worth it simply for the reduction in drunk driving.
Drink driving laws have had a huge negative impact on country living in rural Ireland, where the main social outlet is the local pub, and there isn't a taxi in the village.
This is just a matter of the practical design choices. Any machinery that is in heavy use usually gets fitted, appropriately, with more robust set of parts which last many times longer before wearing out.
Current consumer cars have bearings, joints, and moving parts that are carefully optimized to match the expected usage pattern (which is mostly idle) for a designated period of time and nothing more. That's why older cars can sometimes run for ages. Decades ago we didn't know how to make extremely light-weight parts from least amount of steel with a calculable expiry time of, for example, 40 thousand miles so engineers had to fit cars with slightly heavier and more expensive parts to make sure they didn't break too easily. Think about fitting bearings and joints from a heavy van into a light Japanese small car. Or consider old 70's-80's Saabs and Volvos that can last nearly forever.
Likewise for fluids and lubrication, it's easier to clock high mileages with a car that is mostly in use throughout the day rather than with one that is used a couple of times a day for commuting. The engine wearout is at its peak during the first miles after a cold start.