No, the car sharing aspect of robocars isn't as exciting as people hope


Frequently, in reporting on robocars, it is often cited that one of their key benefits will be the way they enable car sharing, greatly reducing the number of cars that need to exist to serve the population. It is sometimes predicted that we'll need to make fewer cars, which is good for the environment.

It is indeed true -- robotaxi service, with cars that deliver themselves and drop you off, does greatly enable car sharing. But from the standpoint of modern car sharing, it may enable it too well, and we may end up having to manufacture more cars, not fewer.

Today's car sharing companies report statistics that they replace around 13 privately owned cars for every car in the carsharing fleet. Some suggest it's even as high as 20.

This number is impossible for average drivers, however. The average car is driven 12,000 miles/year. To replace 13 average cars would require a vehicle that was actively driving, not just signed out, 11 hours/day and each vehicle would wear out in 1-2 years.

Three things are happening.

  • Carsharing is replacing the more marginal, less used vehicles. A household replaces a 2nd or 3rd car. Carsharing is almost always used by people who do not commute by car.
  • Carsharing is often considerably less convenient than a private car. It discourages driving, pushing its users into other modes of transport, or selecting for customers who can do that.
  • Related to that, carsharing shows the true cost of car ownership and makes it incremental. That cost is around $20/hour, and people rethink trips when they see the full cost laid out per mile or per hour. With private cars, they ignore most of the cost and focus only on the gasoline, if that.

The "problem" with robocars is that they're not going to be worse than having a private car. In many ways they will be better. So they will do very little of the discouragement of car use caused by present day carshare models. The "dark secret" of carsharing is that it succeeds so well at replacing cars because of its flaws, not just its virtues.

Robotic taxis can be priced incrementally, with per-mile or per-hour costs, and these costs will initially be similar to the mostly unperceived per-mile or per-hour costs of private car ownership, though they will get cheaper in the future. This revelation of the price will discourage some driving, though robotaxi companies, hoping to encourage more business, will likely create pricing models which match the way people pay for cars (such as monthly lease fees with only gasoline costs during use) to get people to use more of the product.

There is an even stronger factor when it comes to robotaxis. A hard-working robotaxi will indeed serve many people, and as such it will put on a lot of miles every year. It will thus wear out much faster, and be taken out of service within 4-5 years. This is the case with today's human driven taxicabs, which travel about 60,000 miles/year in places like New York.

The lifetime of a robotaxi will be measured almost exclusively in miles or engine-hours, not years. The more miles people travel, the more vehicles will need to be built. It doesn't matter how much people are sharing them.

The core formula is simple.

Cars made = Vehicle Miles Travelled (VMT) / Car lifetime in miles

The amount of sharing of vehicles is not a factor in this equation, other than when it affects VMT.

Today the average car lasts 200,000 miles in California. To be clear, if you have 8,000 customers and they will travel two billion miles in 20 years (that's the average) then they are going to need 8,000 cars over those years. It almost doesn't matter if you serve them with their own private car, and it lasts all 20 years, or if you get 2,000 cars and they serve 4 people each on average and wear out after 5 years.

There is some improvement over today's cars, but less than you expect. Today's cars also mostly wear out by miles. Some things, like those exposed to the sun, wear out by the year, but not too many. You can store and garage a car that is not being driven and it will stay in great shape for decades with very minor maintenance.

One thing that does wear out with time in today's cars is the technology. Thanks to computers and other tech advances, newer cars have more features than old cars, and so are more valued even then low-mileage old cars. This is a boon for robocars, which will be very technology dependent. We'll be glad they are wearing out in 5 years.

The smaller differences

I said it almost doesn't matter. There are some gains to be had by sharing. They are not trivial but they are not the huge win that people hope.

  • There is less capital investment to buy 2,000 cars every 5 years than to buy 8,000 and hold them for 20 years. Right now, capital is cheap, so the total interest cost is just a few percent of the total costs.
  • As noted, your cars have newer technology as the years past -- a vital element for many.
  • You only need to store 2,000 cars, so you need a lot less waiting space (what used to be parking spaces) and very few parking lots and garages. That benefit was already coming from robocars that can park themselves more densely in available space some distance away from the target location, but this makes it even stronger.
  • Thanks to recycling, the total amount of steel and certain other materials tied up in vehicles goes down.

These advantages pair off with some of the disadvantages of car sharing:

  • Potentially longer wait times, and rare situations with particularly long waits at peak times.
  • Hard to keep your "stuff" in the car.
  • Harder to personalize your style using your car, as people like to do.
  • Less ability to make your car be just what you want it to be without increasing wait time.
  • Logistics of billing, management, damage/soiling.
  • Wasted energy and pollution from empty car moves, which increase the VMT.

There is one factor that may alter these numbers. Today cars are built to last 200,000 miles and 19 years. Future robocars will be engineered with different lifecycle goals. Many of today's cars often last much longer than 200,000 miles with more attentive maintenance. Cars designed to reliably last 300,000 or 400,000 miles (but just 5-10 years) might make a lot of sense. Make cars last 400,000 miles and you only need half as many cars, and that's a huge saving, worth paying a fair bit extra for in the cost of the car and its maintenance.

The lifetime of today's cars is a complex beast. One sector of the market sells their car when it is 4 years old. They always want a late model car and will pay for it. Those unwilling or unable to spend that on a car buy used cars and make use of the latter years of a car's life. Some people, regardless of income, are happy to buy one car and wear it out directly.

Often that "wearing out" point comes when the car needs a repair, and this repair clearly costs more than its worth doing on such an old car. Cars engineered for a different lifecycle could be designed to make this happen smoothly and extract more life per dollar from the cars.

A car's interior also wears out as people sit in it. Dirt and damage accumulate, the leather or plastic wears out or cracks. A car with 250,000 miles on it does not have a nice, shiny new interior the way people like. To last this long, cars must be designed with planned maintenance to the interior as well as the moving parts. A car's finish in part ages with years but also with driving. With luck the robocars will face fewer scratches, but they'll still get dinged by stones on the road.

Sharing a ride

Another factor which could reduce the total VMT is to get people to share rides, not just vehicles. There are many companies out there trying to make that happen with regular cars, but success is modest. There is much more potential with robocars for serendipitous ride sharing that minimizes inconvenience. Here, the system would notice two people who literally are travelling the same stretch of road at almost exactly the same time, and arrange a share. This is highly efficient. Unlike other approaches to ride-share, this can work because individual cars can take some of the sharers on the disjoint parts of their route. So you get in a small taxi that takes you to a parking spot somewhere, where a multi-person car has just pulled up. You get in, share the ride until your paths diverge, and then the car pauses beside another single person car that takes you (or one of the others) along their quick path. Everybody travels almost exactly on their optimal route, and inconvenience is minimized.

Such a system could increase average occupancy. That in turn would drop total vehicle miles travelled, and reduce the total amount of cars needed to be made.

There are many other robocar factors that might change VMT, which changes now many cars must be made:

  • Robocars may encourage longer commutes and sprawl. They might also encourage tighter neighbourhoods too.
  • Deliverbots will eliminate many shopping trips. They will be much simpler and cheaper than passenger cars.
  • People will switch many trips from transit to cars
  • Personal taxi service (parent taking the kids to soccer practice) might be reduced, cutting miles for such trips in half. On the other hand, children may travel more where today they must beg a parent.
  • Some really long trips (like sleeper cars) might arise.


At least with today's car designs, robocars will not reduce the number of cars made as long as people travel the same number of miles. They might even travel more miles, and require empty vehicle moves, meaning that more cars are made per year, though only modestly more. Car sharing of vehicles that are designed to survive many more miles than today's cars would reduce the number of vehicles manufactured each year.

A side note: You will note I have often referred to a car's life in engine-hours. We almost universally use miles to calculate wear and tear on cars, but I think hours would be much better, and cars actually do record this in their internal computers. Highway driving at 70mph racks on the miles but is fairly low-stress, and even lower in accidents. Lower speed city driving with starts and stops puts more wear per mile on a car. Hours are not perfect, but they are better.


I think there's another aspect that you didn't mention - car sharing will significantly increase traffic. I don't want to say it's a bad thing - I welcome everything that increases quality of life. But I don't think that car sharing is a way of decreasing traffic.
Some background: I live in a large German city with pretty good public transport. My home and my workplace are in the same city, about 7km (4 miles) apart. I was over 30 years old when I made my drivers license and have never owned a car. For a long time I was quite comfortable using public transport, because I didn't know better. 1.5 years ago we got two excellent car sharing services, Car2Go (Smart) and Drive-Now (BMW/Mini), which allow you to reserve cars via smart phone and park your car almost anywhere in the city.
The effect, at least for me, is that I use public transport less and less. Carsharing not only replaces the occasional use of a taxi, but also public transport. At some point you start asking yourself "I just missed the train - do I really want to waste 20 minutes of my life waiting for the next train, instead of spending a few Euros to get the next car and be home in 20 minutes". Increasingly I just take the car, because it is faster and easier - even when I don't miss the train.

With robotic cars I would expect it to become even more convenient. You don't have to walk to the next car anymore, the car can pick you up. No need to search for parking spaces anymore (the only real nuisance with car sharing, I usually have to park 5-10 minutes away from my home), the car can take care of itself. Robotic cars would probably be able to reduce my daily commute from an average of 70 minutes to less than 30 minutes. That's much better than the 50 minutes it would take with a car of my own (which still needs to be parked somewhere far away).

Yes, there are various factors where robocars move trips from transit to cars (in fact eventually most of them) but also to shared ride vanpools so there are compensating factors. For example, in areas where a train still makes sense, perhaps the train only goes every 30 minutes instead of every 15, and then at 5, 10, 15 and 20 past the hour a vanpool leaves (robotic) so you still share a ride but you never worry about missing it that much. (The main train is still more comfortable and a bit faster with its private ROW.)

Of course the train should also just have one robotic car every 5 minutes instead of a 6 car train every 30 minutes.

If enough people shift from public transit to individual vehicles then at some point the increased density of cars on the road will increase the trip time through gridlock and eliminate any speed and/or conveniece advantages.

Considering that a single bus can hold about 80 passengers, and a double length or double-decker bus can hold 130+, replacing even one of those that's full at commute time with 80-130 additional individual vehicles in a city business district would already have a significant impact.

So regarding moving trips from transit to cars, "in fact eventually most of them" is clearly simply not going to be the case.

Also, right now there are many people who commute longer distances to and from city centers by car. (See: commute traffic on the bridges and highways in the SF Bay Area) Many of these trips can be replaced by taking a robotaxi or shared robocar service from home to transit, and then taking that into the city, with a reverse of that to go home.

In general, roads in populous areas already tend to operate at capacity. We both know how adding lanes or roads doesn't actually alleviate congestion, it only winds up getting more people to use them.

There simply isn't going to be a big shift from transit to cars, self-driving or otherwise, for no other reason than that the road capacity isn't there. Instead there's more likely to be a bi-directional flow between the two, as robocars make some transit trips more attractive (especially inter-city ones), and some less so, like the example intra-city 4 mile trips. Even then, a 4 mile trip with headways of 20 minutes is probably not a very good example of typical transit in a city. 4 miles is only a bit more than halfway across San Francisco's 7 mile width. Daytime headways for that trip here are more like 10-15 minutes.

"Of course the train should also just have one robotic car every 5 minutes instead of a 6 car train every 30 minutes."

So, like BART, with its daytime 10-15 minute headways per line (or 3-5 minutes per train in stations with multiple lines), rather than Caltrain with its 30 minute ones? Although those daytime BART trains aren't single-car trains with only 60-200 passengers each. Instead, they're 4- or 5-car trains each carrying up to ~1,000 passengers during rush hour.

Yeah, let's see those passenger trips all shifted to cars. ;)

Only a very small minority of commuters (outside of NYC) travel by transit. Moving all those commuters to cars is not nothing, but it's in the noise in many towns. With the robocar's ability to increase road capacity via various methods discussed at robocars/congestion.html this is not a big deal.

Per weekday:*

BART carries 373,945 passengers. Caltrain carries 47,060. SF Muni carries 731,400. AC Transit carries 236,000.

The Chicago El carries 788,415. The DC Metro carries 787,123.

Even the relatively new (now 23 years old) LA Metro carries 360,867.

The NYC Subway? 5,380,184.

These aren't the only ones, and every city and metropolitan area has heavy ridership on busses, in addition to their light and/or heavy rail transit.

And that's just the U.S. In Europe and much of Asia there's heavier use of transit systems.

But that's ok, we can just handwave all of those away when "in fact eventually most [transit trips]" are replaced by robocars, simply with half-size vehicles, better routing, and bunching them up more on the roads. (Oh, wait, the cars on the roads are already bunched up pretty closely at rush hour...)

*All figures from Wikipedia. Cited.

Shows a nice chart of what fraction of trips are transit. Outside of New York (the big winner) and a few others it's under 10%

1. The chart on that Wikipedia page (your blog's comment processor damaged the URL, BTW) only showed North America. Is not the U.S. excessively biased towards private vehicle ownership compared to other countries in the world? What are similar figures for other heavily developed regions, such as Europe and Asia? (Unfortunately Wikipedia doesn't have similar simple "Mass transit in Europe," "Mass transit in Asia," etc. pages.)

2. The chart does show Mexico City and Guadalajara, where transit carries about 2/3rds of all commuters as a percentage of the total. Only 16% of the world's population live in first world nations (1.14 billion vs 7.13 billion). I expect the remaining 84% would look much more like Mexico City and Guadalajara in terms of percentage of commute trips on transit.

But I am keeping my focus on US and other developed countries with high car use, that's where most of the driving and transportation energy use takes place, that is where to design the improvements. The US is, best I can tell, a surprising fraction of the world's driving, more than 1/3 according to some suggestions.

It takes you 70 minutes to commute just 7 km? That's only 6 km/hour. Average walking speed is 5 km/hour.

Even San Francisco's much maligned Muni system has an average speed of 13 km/hour (8 MPH), which is more than twice the speed of your commute.

What "large German city with pretty good public transport" do you live in?

Or do you mean 35 minutes in each direction, for a round trip time of 70 minutes? That's still not "pretty good public transport" in terms of trip times for urban transit, that's only just ok.

I don't know, but maybe this is the door-to-door time: walk to the station, wait a bit, get on, get off to change, wait a bit, get on the next one, maybe change again, get off, walk to final destination. The door-to-door time depends strongly on how often one has to change, how far it is from home to the first stop and how far from the last stop to the final destination. These depend on where home and work are.

Still, I prefer it, even if it takes the same amount of time as driving in rush hour: less stress; I can read instead of drive; cheaper (at least in my case), especially when a) one considers all (even just incremental) car costs and b) stop-and-go is the most expensive form of travel by car.

Door to door time is one factor. Predictability is another one -- terrible in the USA for transit, better in Japan. Changes add more delays and unpredictability.

The robotaxi offers door to door. Unpredictability is due to traffic. It's going to be a more comfortable seat with more amenities, and ability to work and read, but not as spacious as a (lightly loaded) train car. No standing or going to the bathroom without stopping, typically.

Rush hour driving will have little stress with the software doing the work, other than over being late.

In an electric car, stop and go is much better, though still not too efficient. For this, I push for metering of the roads to avoid stop-and-go.

Gross generalisation. There is a spectrum from NYC to an exurb of Houston via small-town New England, and the next 50 years will see the more walkable places flourish and the exurbs crumble. See the blog Strong Towns written by a midwestern fiscal conservative, not some biker-hippy.

Robotaxis will eat away at private vehicle ownership from the margins, hastening the end of the exurbs and the re-urbanisation of America. Sprawl-induced congestion is a geometric problem.

From my perspective, self-driving cars are a boon for encouraging diversity of travel modes.

People who own cars right now but don't use them very much might feel that the increased reliability/availability of robotaxis or car share services, and using them to connect to other transportation modes like rail/ferries/etc., will be enough to give up or retire that car, and stop paying for its parking/registration/insurance/fuel/maintenance.

People who commute to work by car along corridors that are also served by other modes (e.g. Caltrain on the peninsula, BART along the east bay, Amtrack elsewhere, etc.) may find that self-driving car service finally makes it easy enough to get to and from the stations that they stop making the 30-60 mile drive in their personal automobiles.

As it is, train stations are frequently in inconvenient locations, and they often have large park-and-ride parking lots or structures attached to them. Robotaxis or car share services can balance out the disadvantages to taking public transit to a station, or paying for parking and the other associated costs of owning a car and parking it in the park-and-ride lot all day.

People who don't already own a car will have a more reliable means of doing things like taking a bus to a store, and then a self-driving car back with their purchases such as groceries, furniture, hardware, etc. I know a lot of people here in San Francisco without cars who have to call upon friends who have cars or trucks to help them get larger purchases (or even "free, as long as you can pick it up" items) back to their homes. That reduces four car trips (friend's house to your house, your house to store, store to your house, friend going home from your house) to just one (store to your house) if you took a bus to the store, or two if you take a robotaxi or shared car service vehicle each way.

Airports have large long-term parking lots or structures. Right now there are a lot of tradeoffs between taking a taxi or shuttle to the airport and taking your personal car and parking it in long-term parking. Robotaxis should split the difference between them, being as convenient as a personal vehicle and cheaper than a taxi or shuttle because it's just another trip to a robotaxi, not a special airport trip. Yes, extra airport trip fees could be added by the robotaxi owners or by the airport, but those would be so transparently gouging that I think it would be difficult for them to get away with it in the face of passenger's wrath.

As I've mentioned here before, self-driving cars and high speed rail are a great combination. The rail trip is faster than driving and more comfortable (you can get up and walk around, *there are bathrooms onboard*, there's an onboard dining car, you can buy and drink alcoholic drinks, etc.), plus cheaper than flying, with less security and baggage handling hassles. At either end you use robotaxis, or self-driving car services as if they were rental cars. Plus the train stations are more centrally located than airports, and you avoid the possible congestion of driving into a city from the outskirts (isn't it fun to drive to L.A., hit that L.A. traffic, and have to spend another 1 to 2 hours in traffic getting to your destination?).

But frankly, it's hard to see why you would do much of the 3-leg car-train-car trips. The only way those could be better than a pure car trip is if there's so much congestion that the train's private ROW gives it a huge advantage (even with stops) or if the trip is long enough that the comfort of the large vehicle is important, or the trip is long enough that the cost of a private vehicle is something you would notice. In the latter case, shared private vehicle (dynamic carpool) usually solves the problem better.

With congestion, we might solve that with metering, but otherwise, if you can give the train a dedicated ROW to make it more attractive, why not just give that more efficiently to vanpools? They are almost as efficient (often more efficient) as trains, are much more frequent and much more flexible on where they go.

Big trains only solve a very limited problem. They are huge and heavy so take a lot of energy, being highly inefficient except when full. They are only full at rush hour or if you make them infrequent. They have to start and stop all the time to pick up people. The less they start and stop the less good they are for people not near the express stations, and the more they start and stop the more annoying, slow and inefficient they are. They only go along fixed routes, and can't easily pass one another.

All these horrible problems for the economies of scale that come at rush hour. And trains that only run at rush hour are efficient. But right now that's rare, and so they run them mid-day. Robocars can fix that, because you don't have to assure people who ride the train in at rush hour that there are trains back all day, they can take a car off-peak. But then it's a lot of investment just for the peak trips, and the ROW is wasted for most of the day.

The only way those could be better than a pure car trip is if there’s so much congestion

You're writing as if this congestion doesn't already exist.

Roads are already at capacity during rush hour, making road travel slow, regardless of whether you're driving yourself, or a computer is driving you.

Tens of thousands of people flow in and out of central business districts in every major metro area every weekday. Many of them who drive right now would likely be interested in taking faster transit into the downtown area if the outlying stations were more convenient to them.

As it is, there are huge numbers of park-and-ride lots already, where people drive to a station, park, and then take transit.

You said "it’s hard to see why you would do much of the 3-leg car-train-car trips," but your answer is right there. People are already doing it. Robocars will actually make this even more attractive. They will also enable more people to do the same kind of thing for stations that don't even have park-and-ride lots, as in more urban areas.

And it's not necessarily a "3-leg car-train-car trip," especially when traveling to/from city stations downtown.

if you can give the train a dedicated ROW to make it more attractive, why not just give that more efficiently to vanpools?

This is as if it's a question of building train systems, and an alternative is replacing them with vanpool ROWs. But many high capacity transit systems already exist. There aren't as many in the U.S. as in Europe and Asia, sure, but we shouldn't just be U.S.-centric, here.

(Ok, one last one. This was going to be part of my first reply in this thread, but I split it up as it had a separate focus.)

One of the drawbacks I see of existing car sharing services is that you first have to get from your home to where the vehicle is parked. This could be several blocks away. If you need to carry something (like luggage) it might not have been convenient to take it with you on your trip to pick up the car, so you'll need to then drive home to pick it up. Then your home may be in the opposite direction from your actual destination, so you're wasting time and energy retracing your steps. Of course the same problems exist in reverse when you return home.

Additionally, existing car sharing services make you pay for the entire length of time you have a car, so its not economical if you're going to go somewhere and stay there for a while (the movies, a night out, a friend or relative's house, etc.).

A car sharing service with self-driving cars should help solve these problems. The car can come to you and pick you up at home. You can release the car to serve other users if you're not just making a quick round trip, or need to store something in the car. Getting home you can be dropped off at your doorstep and the car will go return itself to its parking spot or be called to serve another user.

I personally think car sharing services are likely to be the most economical use case for self-driving cars. You'll pay a monthly fee to be a member, but then your per-mile trip costs should be much lower than non-membership robotaxis.

Considering the opposition to robotaxis that's surely to exist at the start from taxi companies and taxi drivers, car sharing services are likely to be the main way that non-individually owned self-driving cars are initially available to the public. ZipCar and CityCarShare will probably jump on board quickly with self-driving cars. Although I suppose that Uber and other companies that are trying to disrupt existing taxi services will probably also find self-driving cars a good fit, too. Then there's services like Lyft, where the cars are being driven by their owners. In a similar way they (or some other company) could organize a service to allow people to rent out their personally owned self-driving cars at times that they're not using them. There's already at least one other company allowing people to hire out their cars when they're not using them, as kind of an auto sharing version of AirBnB, though I can't recall the company name.

Finally, studies show that younger people are driving less. So it appears that owning and personalizing a car are not as much of an incentive to them as for older people.

I can see why the taxi drivers would oppose the robotaxis. But why on earth would the taxi companies oppose them? They will be saying "shut up and take my money," I think.

Car sharing companies (some of them) do charge you by the hour of occupancy, including being stopped. They have to as a car really costs $20/hour to operate and they rent it for $9/hour. Some new car share companies let you rent one way, and their costs are closer to $24/hour, a modest profit. One charges 38 cents/minute ($22.80/hour) but caps an hour at a lower rate, because truth is, almost nobody uses these for one way drives in the 40 minutes to one hour range one way. Drive longer than an hour and you can't be one way so they are getting lots of money from you.

Ideal pricing would be around $25 per engine hour and a very small charge like $1/hour for stationary holding. Truth is that the cost of the car sitting there is quite low, but the opportunity cost is actually higher.

why on earth would the taxi companies oppose them?

That depends on classification, licensing, regulation, etc. More specifically, what they can use their influence to try to get enacted. Taxi companies are historically short-sighted and protectionist. They may misguidedly try to push through legislation that forces an official classification of "taxi service" (i.e. waiting at ranks and/or flagged down on the street) as a vehicle that requires a human driver, in an attempt to artificially protect their business against "upstart" companies and technologies.

It's not like the past isn't littered with examples of industries fighting against inventions that would actually benefit them if they just chose to adopt them.

The rose-colored view of technologists to advancements and disruptions is not necessarily the same view held by most of the rest of the world.

The old companies could be short sighted, but the biggest users of these cars will be taxi companies, of a new sort. The old ones might want to block the competition, but they also will salivate over not having to pay drivers or depend on them

That's a good point that a rob-car reduces the hassle of having to go to the shared car to use it.

However, it's still not as nice to use a shared car as a personal one, even if the shared care has KITT levels of self-driving support. You don't always know when you are going want to go out the door and go somewhere. Any time you fail to anticipate the exact time you will go somewhere, you are going to end up waiting for the robo-car to drive itself over.

The only way I see shared cars being beneficial is to save money. Leaving aside the money, they seem clearly worse.

The only way I see shared cars being beneficial is to save money. Leaving aside the money, they seem clearly worse.

"But aside from that, Mrs. Lincoln, how did you enjoy the play?"

For the average person, financial incentive tends to be significant.

There are continuums of convenience and value across owned vehicles, shared car services, and robotaxis. The convenience of a shared car service should be roughly equivalent to robotaxis, but cost less. Owned vehicles are more convenient, but their costs average out significantly higher, depending on the owner's total usage.

Quite a number of shared car services exist right now. Many people find them worthwhile even with the drawbacks that they have. With self-driving cars a lot of those drawbacks go away, making those services even more attractive.

Also, if you don't have your own parking spot (like many people in cities), you're going to have to wait for a robocar that you own to drive over to you from where it parked itself, anyway. So there'd be no advantage in that regard over using a shared car service.

Surely once the older, younger and disabled start using robotaxis the overall miles traveled will increase. A quick look at some stats in US 2010 there were about 209M licensed drivers with a population of about 308M. 23.5% of the population was below 18 or 72M. 308M - 209M - 72M = 27M+ adults without a license. There were 20M kids between 10 and 14 years old and 22M between 15 and 19. robotaxis are looking at serving 30M-50M new people possible even more that currently rely on someone else or transit for transportation. That is a 14% - 24% increase. The roads are going to be packed with robotaxis. I think the small tandem 2-seater robotaxi will be important to alleviate congestion and make robotaxis affordable to the masses.

Thanks for this copious response to my robotaxi suggestion in the previous post.

* The re-urbanisation of America is finally underway. The generation that grew up in the back seat of the SUV is shunning the cul-de-sac, congested dendritic street network and separated uses of suburbia for a grid of narrow, slow streets dominated by foot and bike traffic and a pleasant walk to the corner store and transit boulevard. This is not about peak oil as much as it is about the realisation that the post-war suburban experiment was an historic anomaly, completely out of kilter with the past millenia of human-scaled development patterns and basic human psychological needs and desires.

* With a robotaxi service on offer, not owning a car (or going from two cars to one) becomes a more viable proposition in more places. At the margin, this makes retrofitting suburbia easier. You noted you have to wait for a robotaxi to come: this will encourage more walking, at the margin, and increase demand for walkability retrofits. (See the books Retrofitting Suburbia and The Sprawl Repair Manual).

* You noted that robotaxis, by being used more efficiently, will be more modern than owned vehicles. This will make them more attractive. Why lease on something that will be out of date quickly, when you can just pay-per-use for the latest?

* I can't see any of the disadvantages that you list as effecting anything other than a niche of the market.

* I think we agree that path-dependency is important, and therefore I expect most people's first contact with auto-autos will be as robotaxis, since they'll be expensive. Coupled with the timeframe for re-urbanisation, I don't expect there ever to be a time when auto-autos are both cheap and desirable enough for enough people for personal ownership become a significant part of the auto-auto market. It'll be robotaxis to replace carshare and taxis on non-transit routes in town (and for non-transit/train inter-urban routes) and human-driven vehicles for sparse rural areas.

I have an article on the potential for re-urbanization at you might find of interest

American cities aren't filled with parking craters because Americans (exogenously) like to drive and park a lot. Instead Americans are forced to drive a lot because the cities are so unwalkably sparse: they're not designed at a human scale.

They're so unwalkable because of car-first engineering street design standards (AASHTO's green book) and because of zoning (for single use, setbacks, parking minima etc.)

You could fix both of those regulations today - and many cities are - and start building on the wide streets, making them thinner, or 'complete' with dedicated transit and bike lanes. You can also start building on the empty surface lots, by fixing your zoning codes. Infill is all the rage. Transect/form-based codes are steamrolling across the country. The New Urbanism is now the establishment in many places.

So leading cities, with historic grids, are already infilling. My claim is that robotaxis will allow the more marginal cases to join this trend. That's a different claim than saying robotaxis will spark a reorganisation of cities.

Technology makes regulatory changes more politically palatable. But it's the regulations - which can be made without the technology, if they 'return' us to the pre-suburban experiment development paradigm - that define the cities.

"One of the most fundamental purposes of the city is transportation." - Not quite: the purpose is access. Transportation is a means to that access, but can be vastly reduced to the point of being unnoticeable if the things to be accessed are sufficiently close together. The walk to work - if it passes by friends' houses, grocery stores etc. - is about more than transport. The bike to work, if it's pleasurable, is about more than transport.

Time is really important to people: the normal curve centers around a 30 minute commute, no matter how productive or pleasurable, or by what mode.

"The car itself caused cities to change in the 20th century and drove suburbanization and other trends." - Again, not quite. The city design laws (AASHTO and Euclid) drove suburbanization. Also massive Federal subsidies for road building and insurance only for single-use, sprawl-type mortgages. Mixed-use Main Street USA doesn't securitize very easily.

I highly recommend familiarising yourself with the Urban Transect. It allows a lot of clarity in urbanism discussions. The various forms of density you describe (main street, suburban single family, downtown) all have their place. The key is to avoid monotony, and ensure the streetscape is pedestrian-first: short blocks, thin or complete streets.

Hey, Brad -
Just as the freeway presented a natural progression, wouldn't it be 'natural' to have robocar meet a different kind of freeway: guideway? Robocar drives you from home to freeway, except instead of entering the on-ramp for the freeway, this on-ramp seamlessly converts your ride from terrestrial to extra-terrestrial - an elevated guideway. Just like the old-style on-ramp, Robocar quickly accelerate to guideway speed (100 mph) and off you go, the vehicles platooning so that a single lane now handles the capacity of 20 former lanes. And remember, lanes can be stacked. When you get close to your pre-programmed exit, the platoon opens up and you switch to the off-ramp, while everyone else goes about their merry way.
The only diff is, this robocar would have a mating means (either top (roof)-mounted, or bottom mounted. Good renderings and description via Roger Davidheiser's Third Generation roadway.
Two neat things about it. First, being on a smooth guideway, the wear and tear on the vehicle is dramatically reduced, hence greater longevity; and second, there is no surface intrusion whatever. The only thing you'd have to worry about would be collisions with birds.

Of course the issue is these guideways don't exist, and even if you built them, they would only exist in the few places you built them. So that's not particularly interesting. Useful as a means to increase capacity and provide guaranteed trip times in a few highly congested areas, but unlikely to be used outside of that.

Convoys and mating are things that might happen in the more distant future, but the first cars on the road have nobody to talk to or mate with, so you don't solve that one first, not even 5th.

The actual average number of miles driven per year - per car is 4800, not 12000.

There are about 1.2 Trillion Vehicle-Miles-Traveled each year:

There are about 250 Million vehicles:

Therefore, there is additional scope to reduce the fleet 60% beyond what you anticipated above! That is an enormous amount and must change your conclusion.

Can you be more specific about what item on the FHWA page led you to think there are only 1.2T VMT? The number is actually just under 3 trillion.

Arg - you're exactly right. I misread cumulative for annual. Sigh.

I find it difficult to make predictions, especially about the future. So, it is useful to explore all options.

One that I find interesting is car clubs. In this case, a person would belong to a club that guaranteed them a car of particular characteristics, quality, and condition at specific times (say rush hour). Then, they would have the option to use at other times and the cars could be farmed out as robotaxis in the interim.

These clubs could be mutual organizations, or subscription services by for-profit companies.

The club concept could address several issues:

  • Guaranteed availability at high-demand times (robotaxis may be in short supply at rush hour, for example - reducing uncertainty has value).
  • Guaranteed quality and condition of the vehicle, as opposed to robotaxis which will always be more uncertain than private vehicles.
  • The high-capital costs of robocars could be amortized over more people and more time.
  • Robotaxi operation offsets the cost of ownership.
  • The organization standardizes the process for robotaxi operation.
  • The organization could have facilities for cleaning and maintenance to ensure quality after robotaxi operation.
  • The club-member has no need to have facilities at home to accommodate the vehicle (i.e. garage, charging station, etc.).

I can't guarantee this will be a popular option, but it should be considered as a compromise between the models of private ownership and robotaxis. It is similar to today's car-sharing, but potentially offers some additional benefits.

Yes, I'm a big fan of this model and written about it a fair bit before. In my week of robocars I describe something called the "Red Beemer Club" that always gets you a red BMW, allowing people to still get identity from cars and to buy into brands, something they won't do as much with general taxi service.

If the club hires out its cars, its just a variation on individuals owning a car they hire out. Right now, at least, capital is cheap, so it is not a big barrier to getting any sort of fleet, whether small and private or larger.

Once your fleet gets to a certain amount of usage, your cars start aging by the mile and what aging they do by the year becomes less important, and as such, capital becomes less important. If you want to own a car solo, then your car ages by the year and capital becomes a bigger part of the equation.

Being in a club (or private owner) can offer greater assurance of a car, but so can taxi fleets that offer service level agreements. Strictly, the latter are able to do it more efficiently. What they would not likely do (though they could) is offer you the same style of car.

However, one big question is this. When you want a car now with no advanced notice, you will find yourself having a choice of a particular class of car in some number of minutes, or the "first available car" in less time. Will people choose to wait for the car of choice, or just take the first thing that can come get them?

To your question: people may simply choose the first available when a random need appears.

However, for regular needs like commuting they may very well may choose one more appropriate to their needs or wants.

Also, very likely the fleet size will be significantly smaller than now (my above error notwithstanding). The supply will be significantly constrained in rush hour, especially in areas where the origins are substantially separate from the destinations. For example: a case where most of the people live in outlying residential areas while the jobs are in a downtown core. There will have to be someway to deal with this effect: dramatic congestion pricing, flex-time, etc. Some people will need access to a car no matter what. A club could likely be much more attractive than private ownership.

The supply of cars will be up to the market. If the market wants a car for everybody at rush hour, and will pay for it, that's still far less cars than are present today. The more cars in the fleet, the longer the turn-over time in the fleet, but otherwise costs are only marginally higher for a larger fleet. The benefit is faster response time from the larger fleet.

All these questions the market will be answering, including how much people just take the first available car, and how much they want their specialty car.

Yes, of course, "appointment" trips will always get the car desired if it is available when booked. One of the nice tricks is you don't need a large fleet to meet customer's appointment trips. Only immediate response benefits so strongly from a large fleet. I see many people using one service for their appointment trips, and "whichever service in my price range can get me a car soonest" on their spontaneous trips. But even 10 minutes warning changes that equation, and most trips have 10 minutes warning I think. (If people are up to doing that.)

Brad, this reminds me of another concept that needs some holes poked in it. I often hear, "With autonomous vehicles as taxis, you can commute in an efficient vehicle during the week, and only grab that SUV or minivan on the weekends when you need it." I like it, except I've been there. I was a 1-car family for a while going under that assumption. Except, the days I wanted the large vehicle were the days everyone else also wanted a large vehicle. Ski season, holidays, Stanford graduation weekend.

Rental car companies aren't structured to keep a high enough level of inventory so that people have cars whenever they want them. There is some minimum level of weekend activity they need for it to be profitable to keep a car in the fleet, and when the demand varies, some people aren't getting that vehicle.

Maybe I'm wrong. Maybe someone who understands the overall economics of robotaxis could find it optimal to keep "unprofitable" cars in stock for the weekend because they will be making profit through the daily use of the other vehicles. I'd love to hear your thoughts.

Having enough SUVs to meet peak SUV demand at the height of ski season or a holiday is still a lot less SUVs than we have now. Plus you can have the SUVs operate only in the mountains -- take a regular car to the mountains, switch to SUV when the terrain requires it. At least if the regular car has a ski rack. Then -- though it is somewhat wasteful -- the SUVs can shuttle back and forth and you need fewer of them.

The extra big vehicle can always be used to serve a single person. It's just not efficient. But today, every SUV user taking the SUV to the grocery store -- that's not efficient at all.

"only grab that SUV or minivan on the weekends when you need it"

I suspect that close to 100% of people who drive an SUV never actually need it.

SUVs are to sensible cars what bling bling rap/hiphop jewellery, grills etc are to the crown jewels.

Switzerland is seriously considering banning them altogether.

Conspicuous consumption is fine, but it has to stop when pollution is much too much and the safety of others is endangered.

You might argue people rarely need the SUV. (It's not never, I've really gone off-roading, really used a borrowed SUV for towing etc.) But people want it. It's high up, roomy, powerful. As long as they want it, they will buy it even if they rarely or never need it. But if they can get it when they need it and pay less the rest of the time, many will be happy with that.

> We almost universally use miles to calculate wear and tear on cars, but I think hours would be much better

The amount of work (in Watts) can be used for a vehicle lifetime optimisation. Yes, that is not very representative to the end user, but it can be an essential parameter for estimation of vehicle sharing efficiency. E.g. for future shared vehicle network simulations and efficiency monitoring.

Aside from the fuel energy conversion efficiency (which is calculated in all modern engine control systems) it consists of two parts:
= work to transport useful load / passenger(s) - this part should be increased in optimisation
= and work to be done to move the car itself - this part should be decreased

A car is of course a complex thing and parts wear out from different things. Watts could be good, and gallons of fuel would of course correspond quite closely to that, but the reality is a complex series of things:

  • Total engine revolutions or power, as discussed
  • Operation while cold plays a very disproportionate part
  • Starts and stops affect both the engine, and the brakes
  • Hours of occupancy affects wear on the seats, the wheel, dirt etc.
  • Number of bumps and shakes will affect both suspension and many other things in the vehicle
  • Stress events -- fast starts, hard stops, hard turns, hard bumps, bad fuels / dirty air, cold starts -- might be a good thing to measure.

If we measured those (and most cars can measure most of these with the odbII port and perhaps a phone with accelerometer) we might find out whether gallons, hours, total revs or miles is the best proxy.

> A car is of course a complex thing and parts wear out from different things. Watts could be good, and gallons of fuel would of course correspond quite closely to that,

Sorry my fault. I was meaning - work in Joules :) which has to be made to move the car from point A to B with or without useful load. Of course if the car is stopped it does not make any work :) so the power in Watts (as time integral of potential energy change) is also OK.

With introduction of parallel hybrid (range extended) and full electrical cars chemical fuel becomes not the only one source of energy. Using a work in Joules in vehicle operation efficiency calculation is.. how to say.. more fundamental. And today there is enough computing power to calculate in real time.

[Chemical power] + [electric charge power] = [wasted power] + [power to move the car itself] + [power to move the passengers/useful load]

The idea of introducing useful work as a parameter of optimization lies more in the field of centralized traffic optimization. But it also can be used to estimate the lifetime of the car's powertrain. Car driven in the harsh environment (offroad, cold, mountains) clearly consume more energy. Extra energy is used to heat up the engine in cold conditions and to move the car uphill. If we know the amount of energy consumed by the cars in town, we can calculate a desired load on electric grid (this will become more important with more electrical cars on the road in the future) and the required amount of chemical fuel by the car-sharing company.

Shared cars have advantage over privately ones if they are used more effectively. Their driving profile can be optimized. The ratio between the energy used to move the car alone and the car with useful "payload" can be used by a centralized computing system to decide where (e.g. in which district) the cars are more needed in the moment and where the cars should change their "tactics" and their routes to increase their efficiency.

> but the reality is a complex series of things:

Total engine revolutions or power, as discussed
Operation while cold plays a very disproportionate part
Starts and stops affect both the engine, and the brakes
Hours of occupancy affects wear on the seats, the wheel, dirt etc.
Number of bumps and shakes will affect both suspension and many other things in the vehicle
Stress events — fast starts, hard stops, hard turns, hard bumps, bad fuels / dirty air, cold starts — might be a good thing to measure.

This information can be collected by each of system's electronic control unit independently - in a form of wear ratio from 0 to 100%. Then this information has to be summarized by e.g. body control unit and sent to the vehicle control network. After processing it should be clear which car needs maintenance right now and which system exactly has to be repaired or replaced by workshop.

> If we measured those (and most cars can measure most of these with the odbII port and perhaps a phone with accelerometer) we might find out whether gallons, hours, total revs or miles is the best proxy.

But unfortunately on this moment a standard OBD2 functionality (as its required by the norms) is not capable to measure a wear ratio (as a single parameter), but the customer specific software inside vehicle control units are doing that. For example: in the most of modern diesel engines an oil by fuel dilution is being calculated. I also remember a case when our customer asked us to implement a "BlackBox" function that counts a total duration (hours) of engine overspeeding. For warranty of course.

In any case shared cars (robotic or not) require more remote control on a vehicle status than a privately owned one. And I think currently there is no norm for that defining which data should be gathered and how it should be done.

With the combination of what you get from ODBII -- which includes the tachometer reading, and often fuel use readings, as well as the GPS and accelerometers in a standard phone, I think you could measure most of what matters.

Sure, the fuel consumption data is available in OBD2 protocol (unit: liters per hour) but it may be quite unprecise depending much on data requests frequency. The best option is to get the fuel consumption from the CAN bus directly (e.g. by wireless adapter) by reading standardized messages. The most precise data can be read from engine control unit itself - every milligram of injected fuel is being measured on every combustion cycle and integrated into total fuel consumption over all mileage..

For a hybrid vehicle estimation can be a little tricky, while standart OBD2 protocol offers only the percentage of remaining battery charge. So the measurement method has to be adapdet with the car model / battery capacity. Also the actual amount of electrical energy will be unkown to the user, unless a kind of interface (wireless of course) between charging station and user's phone (sending data about how much enegy is being charged) is available..

Add new comment