Will autonomous cars be more economical/efficient as shared urban transit than busses or trains, and by how much? What's some good research on this?

post by MakoYass · 2019-07-31T00:16:59.415Z · score: 10 (5 votes) · LW · GW · 10 comments

This is a question post.

Contents

  Answers
    8 shminux
    2 Dagon
    2 greylag
None
14 comments

There's reason to think they would be. When people get into a shared public transit vehicle, usually, they're not all going to the same place. The vehicle has to take an indirect route, it has to start and stop many times along the way to let people on and off. A person will often have to transfer between multiple lines along the way and that will often involve a bit of waiting. There are usually many routes that have very low utilisation rates- carriages will be mostly empty most of the time, in part due to the fact that you can't deploy half of a bus for low-use routes, in part due to the rigidity of the route scheduling.

Small self-driving vehicles, instead, can take a person straight where they need to go. The vehicle can then pick up someone else nearby and take them straight where they need to go, and so on, all day. They'll also benefit from economy of scale of production, if more units are produced, the cheaper they individually become, smaller units, then, can be cheaper.

That's the vision I have in my head [LW · GW], anyway.

The main potential issue that I can see is commuting patterns. If a lot of people are going in the same direction at the same time, batching their trips together with larger densely packed shared vehicles might make sense.

But I'm not sure that's a legitimate way for cities to be. With affordable inner-city housing, mixed-use planning, this commuting pattern wouldn't pop up - people would be taking short trips in all different directions even during peak commuting times.

(Although I cannot personally understand why) A lot of people seem to like living in suburbs, and conventionally planned cities have land pricing issues that make living in them prohibitively expensive for most ordinary people who work in them (I will discuss a potential solution to this in a later post). These ideal mixed use cities that I am mentally situated in, might not exist, and might not come to exist before autonomous vehicles start to compete with existing public transit.

So I'd like to know what relationship suburb size has to the commuting pattern problem - how large can the suburbs get before fixed-route busses become more efficient? (and also, are they really better, even with large suburbs? Is there that much route agreement in any city?)

Visions of self-driving cars currently look like variants of the cars we have, improved for not needing to have a driver's seat and forward visibility, often you'll see 8-seaters that're expected to be shared to some extent. I think that might be unrealistic, if it is, we need to see more analysis of single-occupant cars like the toyota i-road. Aside from using fewer materials and using less energy during acceleration (they're lighter)[1], they'd also have the advantage of being able to to drive two to a lane, reducing congestion.

If we can answer these questions well, I think we will be much better informed about the future of urban transit than we are now.

[1]: although, regenerative braking is a thing. I'm not sure there's really an upper limit on how efficient that can be made, so maybe weight isn't so much of an issue.

Answers

answer by shminux · 2019-07-31T02:25:23.780Z · score: 8 (4 votes) · LW · GW

I remember thinking through the potential evolution of autonomous transportation* some 10 years ago, and, barring the protectionist forces winning out and enshrining the "right to drive" in law, like the "right to bear arms", it's pretty clear where the transportation is going.

1. One- or two-person electric commute vehicles dominating city traffic, eventually leading to the whole swaths of urban areas being closed to human drivers, which would be deemed unsafe. Those areas will then expand outwards and merge, eventually spreading into the suburbs, and at some point major highways, first with HOV lanes, then taking over the rest lane-by-lane. Owning a car will become very expensive, and a human-driven car prohibitively so.

2. The huge parking lots will disappear, since uber-like electric commuters will be in use much more often and can be stored efficiently in much smaller spaces during off-peak times.

3. Everything will be routinely recorded, whether outside of the vehicle or inside it, limiting the type of activities one can indulge in while getting to the destination. Vandalism will virtually disappear, as well. Ride sharing will complete strangers will be as safe as walking along them on a busy street somewhere in central London.

4. Once there are no more human drivers, in the autonomous-only areas the vehicles themselves will be able to communicate and coordinate, and soon will be required to do so, forming a driving grid. Any vehicle not complying with the grid inclusion rules will not be allowed in, or forced to stop and get towed outside. Yes, the grid will eventually take control from the single vehicles.

5. Once that happens, the traffic lights will be largely obsolete. There will be pedestrian crossings, with the Walk/Stop signs, but no usual traffic lights, since there will be no human drivers to look at them.

6. The current alternating pattern of driving through the intersection will change: without pedestrian crossings cars will simply zoom in all directions, their movement perfectly choreographed by the grid. With pedestrian crossings there will be breaks for humans to cross on foot in all directions at once. Odds are, many crossings will be replaced with walkways above or below ground.

7. Congestion will be greatly reduced due to coordination. Worst case you'd have to wait to get your ride, as the grid will limit the number of vehicles to keep the system at peak efficiency. Traffic jams will be extremely rare, since most current causes of it will be eliminated, such as broken down cars, accidents, high volume traffic, power outages, road work (the grid will shape the traffic around any roadwork in progress).

8. The city architecture will change to accommodate the new transportation realities: there will be much less road space needed, so some of the wide busy streets will be repurposed for parks, living spaces, etc.

This is as much as I recall offhand, but there is definitely more.

Now, to answer your question, the costs will eventually go down orders of magnitude compared to the existing means of transportation. Which does not mean that price will go down nearly as much, as everything will be heavily taxed, like train and plane tickets now.

Edit: I expect this will happen first in places with high penetration of autonomous vehicles. Places like, say, Oslo. Also in the countries where the government can exert some pressure and ensure compliance and coordination, like, say, in China and maybe Japan. The US will be one of the last ones, and the most expensive ones, as is customary with most technological innovations lately.

_________________________________________________________

* The language will evolve accordingly:

  • "self-driving car" will be a name for DIY driving, the opposite of what it is now.
  • self-driving will be reserved for antique car enthusiasts, who would tow their cars to a "driving range" and show off their skills in this ancient activity, sort of like horseback riding is now.


comment by ChristianKl · 2019-07-31T06:51:18.576Z · score: 2 (1 votes) · LW · GW

A general experience of building more roads has been that it causes people to travel more and thus use up the capacity of the new road without relieving the old roads.

I would be very surprised if driver-less cars will result in us needing less streets. We can repurpose parking space but it's unlikely that we can do the same for roads.

comment by AnthonyC · 2019-08-01T15:25:23.126Z · score: 1 (1 votes) · LW · GW

Yeah, I'm not convinced in either direction that the efficiency boost from autonomous vehicles will be able to overcome Smeed's law ( https://en.wikipedia.org/wiki/Smeed%27s_law ).

I still think that trains and subways will make sense for major arteries, and buses for very common (especially not-straight-line) routes.

I do think mixed-use-planning would both reduce trips and increase the impact of autonomous vehicles by making longer-distance commutes less common. However, there's a lot of friction from other parts of society. It's easier to change jobs (and job locations) than to move, especially if you own instead of rent, and especially especially if you live with a partner and you each work in different places in and around the city.

comment by greylag · 2019-07-31T07:05:46.304Z · score: 1 (1 votes) · LW · GW

Autonomy can allow for higher density by:

1) at worst, reducing, at best, eliminating, the headway between vehicles that's needed to allow human drivers to react (@shminux's "zooming in all directions")

2) in busy times and locations, aggregate multiple journeys into multiple-occupancy vehicles running ad-hoc routes. (I think that's what the OECD "shared mobility liveable cities" study is proposing; UberPool is similar; Citymapper's "smart buses" are similar (though all with human drivers))

comment by ChristianKl · 2019-07-31T10:20:49.528Z · score: 4 (3 votes) · LW · GW

Whether or not you get multiple-occupancy vehicles depends on market economics and not what a city planner who wants low traffic desires.

Poor people might go for multiple-occupancy vehicles but I would expect that richer people do want to get faster to their destination.

Getting faster to the destination however isn't the only thing worth consuming. You also want to spend the time in the vehicle well.

It might be that a smart startup figures out how to have multiple-occupancy vehicles that provide a desirable experience of social interaction between the passengers but in the absence of that you can do a lot more for the experience in custom designed vehicles.

You might get a new haircut on your way to work, exercise in a driving gym, get a massage, take a shower or engage in a variety of different experiences that are enabled by specialized vehicles.

The more specialized vehicles you have, the more often you will have vehicles that drive empty to their destination.

answer by Dagon · 2019-08-01T16:21:28.700Z · score: 2 (1 votes) · LW · GW

In the short-medium term (next 30 years, say), change is going to be gradual. It will take a LONG time to overcome all the edge cases of current driver, traveler, and pedestrian expectations. We'll see more ride-share and semi-autonomous, and portions of the ride-share will become autonomous, but not a lot of end-to-end differences.

I haven't seen much study of the demand side of things - what is the actual price/time/convenience elasticity of common travel desires. I know that commute times stay fairly constant - people move further out when roads improve. I don't see why that will change - people will take more distance (and cheaper/roomier housing) over less time/resources for their commute. I do strongly expect that people will simply travel longer and more often as it gets cheaper/easier to do so.

Rail is here to stay, as it combines actual efficiency of resources with the populist demand for central provisioning and subsidies. Busses tick the same boxes (though less convincingly), and probably stay for the same reason.

answer by greylag · 2019-07-31T07:00:35.446Z · score: 2 (2 votes) · LW · GW

@makoyass I think you would be interested in The End of Traffic and the Future of Access . I haven't read it, though I have read some of Levinson's other work; it's a bit on the dry-and-wonkish side, but I expect you would prefer that to "rabid conflict-theorist", and it's covering the right sort of ground.

10 comments

Comments sorted by top scores.

comment by habryka (habryka4) · 2019-07-31T00:19:06.636Z · score: 8 (4 votes) · LW · GW

If I remember correctly, trains have almost an order of magnitude (or something in that space) lower friction than cars or basically anything else, which makes them a much more efficient mode of transportation. The problem is just building rails everywhere.

My guess is that this makes it hard for anything that runs on anything but train wheels to compete energy-wise with train-based public transportation systems.

comment by MakoYass · 2019-07-31T00:59:56.429Z · score: 1 (1 votes) · LW · GW

https://www.itf-oecd.org/shared-mobility-innovation-liveable-cities

examines the effect of replacing all car and bus trips in a mid-sized European city with automatically dispatched door-to-door services. The report finds that such systems can massively reduce the number of cars on city streets while maintaining similar service levels as today. They also result in significant reductions of distances travelled, congestion and negative environmental impacts. Not least, automatically dispatched, door-to-door services also improve access and reduce costs to consumers.
comment by MakoYass · 2019-07-31T00:50:30.262Z · score: 1 (1 votes) · LW · GW

Motivation:

I ask this question in light of seeing this article https://macwright.org/2019/07/27/beware-the-ethical-car.html and getting kind of worried to see these people, again, getting ready to reject a potential economically viable technical solution to an otherwise unsolvable social problem, because they don't believe in such things, because they personally dislike the people proposing them, or because they would deep down prefer a solution that involves humanity atoning for its sins and changing its ways, which they ought to be able to see with their eyes is not something humans do. It is failing to happen in front of your eyes.

So, it may be very important that someone thoroughly answers this question, so that the most viable solution to transport emissions is pursued with the energy it deserves.

It may need positive, constructive political attention to be done really well, it may not be sufficient to leave it to the corporations. Without coordination, it would seem to me that empty superfluences of autonomous vehicles are incented to generate just as much congestion as we have now, jockying for customers by driving around unoccupied wherever they might be. It's also not obvious that it will ever become cheap if a private monopoly manages to take ownership of the user's experience, I'm sure Uber has no desire to share the road with other providers. If some provider manages to secure a monopoly on production, on algorithms, or on licensing... it seems unlikely, but the more disaffected the public are about autonomous vehicles the more likely it is to happen.