Tuesday, August 18, 2015

Preparing for the impact of driverless cars

The buzz has been building about driverless cars for a while now, and this week I want to talk about a couple of new articles on the topic followed by my own thoughts.  The first is a McKinsey article based on MIT research:

Full speed ahead: How the driverless car could transform cities
"By combining ride sharing with car sharing—particularly in a city such as New York—MIT research has shown that it would be possible to take every passenger to his or her destination at the time they need to be there, with 80 percent fewer cars
Clearing the roads of four out of five cars has momentous consequences for cities, by measures such as environment, traffic, efficiency, and even parking. In most cities, for example, designated parking accounts for a huge amount of land, which ends up being useless for most of the day. With fewer cars, much of this space could be freed for other uses. Such reductions in car numbers would also dramatically lower the cost (and related energy consumption) of building and maintaining the roads. One engineering study found that automation could quadruple capacity on any given highway. And, of course, fewer cars also means less noise and a smaller environmental impact. 
Driving patterns of individual cars can be algorithmically optimized as well. Because autonomous vehicles don’t get lost, they create less congestion and shorten travel times. More important, self-driving cars would also make for much safer roads; more than 30,000 people a year die in automobile-related deaths in the United States every year and 1.2 million worldwide."
I do have one quibble with the assertions above: yes, there will be fewer cars, but I suspect there will be a similar number of car *trips* (for example, one taxi providing 20 trips/day instead of 10 owned cars each providing two trips/day), and that means just as much wear and tear on the roads, unless a lot more car sharing happens (i.e. one vehicle carrying multiple people on separate trips at the same time).  More on that later...

The second article is from The Economist and chock full of interesting facts:
  • Cars sit idle 96% of the time.
  • Google thinks self-driving taxis could have utilizations of 75%+.
  • Stanford estimates we'll need 70% fewer cars to provide the same trips.
  • "The idea that autonomous vehicles will be owned and used much as cars are today is a “tenuous assumption”, says Luis Martinez of the International Transport Forum, a division of the OECD, a think-tank. Fleets of self-driving vehicles could, he says, replace all car, taxi and bus trips in a city, providing as much mobility with far fewer vehicles. An OECD study modelling the use of self-driving cars in Lisbon found that shared “taxibots” could reduce the number of cars needed by 80-90%. Similarly, research by Dan Fagnant of the University of Utah, drawing on traffic data for Austin, Texas, found that an autonomous taxi with dynamic ride-sharing could replace ten private vehicles. This is consistent with the finding that one extra car in a car-sharing service typically takes 9-13 cars off the road. Self-driving vehicles could, in short, reduce urban vehicle numbers by as much as 90%."
  • 94% of accidents are from human error, and these could be eliminated.
  • "A study by the Eno Centre for Transportation, a non-profit group, estimates that if 90% of cars on American roads were autonomous, the number of accidents would fall from 5.5m a year to 1.3m, and road deaths from 32,400 to 11,300."
  • "As well as being safer, self-driving vehicles would make traffic flow more smoothly, because they would not brake erratically, could be routed to avoid congestion and could travel close together to increase road capacity. A study by the University of Texas estimates that 90% penetration of self-driving cars in America would be equivalent to a doubling of road capacity and would cut delays by 60% on motorways and 15% on suburban roads. And riders in self-driving vehicles would be able to do other things. Morgan Stanley calculates that the resulting productivity gains would be worth $1.3 trillion a year in America and $5.6 trillion worldwide. Children, the elderly and the disabled could gain more independence." 
  • "With cars in constant use, much less parking space would be needed. Parking accounts for as much as 24% of the area of American cities, and some urban areas have as many as 3.5 parking spaces per car; even so, people looking for parking account for 30% of miles driven in urban business districts. By liberating space wasted on parking, autonomous vehicles could allow more people to live in city centres; but they would also make it easier for workers to live farther out. If you can sleep on the journey a longer commute becomes feasible, notes Mr Fagnant, who foresees a “simultaneous densification of cities, and expansion of the exurbs”.
Again, I think it's worth noting that even though the number of vehicles drops, the amount of vehicle-miles probably stays pretty steady or maybe even increases as people can be productive on longer commutes.  In essence, there will be fewer vehicles, but they will get used up/worn out much more quickly from their high utilization (similar to buses today), so the car industry may be safe from complete collapse, although it will certainly be massively disruptive.

A key question is how much car sharing will occur, which reduces prices and increases efficiency by picking up and dropping off multiple people along routes.  It can be a bit awkward sharing a vehicle with strangers.  I would not be surprised to see someone like Uber custom design a vehicle with individual personal compartments.  Imagine 5-6 private individual seating compartments in a 6-door SUV-sized vehicle.  When it pulls up, an indicator tells you which door to get into for your compartment, and then alerts you again when it's time for you to get out, based on the destination you put into your smart phone.  Private ride, shared prices and efficiency - best of both worlds.  Mass adoption of shared rides would solve our traffic congestion problems almost overnight.

A couple of additional thoughts: If most accidents get eliminated, do we still need shoulders? Maybe those could be converted to extra lanes?  The same for street parking if vehicles are continuously utilized - long-term those spaces might be convertible to additional lanes, adding surface street capacity.  Or in some cases, it might make sense to expand the sidewalk/public realm into that space instead.

So what should cities be doing now to prepare for this future?
  1. Loosen up or even eliminate minimum parking requirements now so available parking starts shrinking naturally over the next few years.  This will also enable greater infill and density in cities as well as supply much-needed new housing stock.
  2. Stop investing in new rail transit - they're not going to be able to achieve their payback before this revolution (if they ever could in any case).  Managed-lane networks are a better investment, as they can be used for buses, HOVs, and toll-payers now, and easily switched over to automated vehicles later.
It's going to be a brave, brave new world...

Labels: , , , ,


At 8:51 PM, August 18, 2015, Anonymous Jay G said...

Depending on how far off the wide spread adoption is, I can see this making the speed train from LA to SF a massive waste. I wonder if speed lanes might evolve for specially designed autonomous vehicles that could go 100+mph.

At 9:00 PM, August 18, 2015, Blogger Tory Gattis said...

I love that idea! LA to SF is a massive waste even without driverless cars, just from the sheer cost. If it were done more affordably, like the Texas plan, it might make sense. But there's also adequate air shuttle capacity between SF and LA.

At 9:04 PM, August 18, 2015, Blogger Tory Gattis said...

I imagine part of the insane cost of CA HSR is making a multi-hundred mile line of very fast trains earthquake proof, because a strong one could kill thousands derailing trains at those speeds. Another reason air shuttles - or cars/vans/buses - are a better choice in CA's case.

At 10:11 PM, August 18, 2015, Blogger Max Concrete said...

I think this transition will play out over a very long time. My opinion is that the transformative impact will be after 2050, maybe well after 2050. The average age of a car in the U.S. was recently reported to be 11.4 years, which implies that the average car has a lifespan of about 23 years. Assuming a 20-year lifespan, this means that new cars sold this year will be on the road until 2035. If self-driving cars reach the market around 2020, they'll start with a negligible market share (similar to electric cars today) and slowly gain share. New conventional cars entering service in the 2020s will be in service through the 2040s. Let's say self-driving cars can get 50% of the market by 2030, which I think is way too optimistic. Even in that optimistic scenario, there will be a high percentage of conventional manual cars on the road in the 2050s, although those cars will likely have some automated capability like adaptive cruise control. To achieve the benefits of self-driving cars, you need a very high percent of the fleet such as the 90% cited in the reports.

I also have not seen any estimated cost of these cars. Self-driving cars will be a technology tour-de-force, with radars, sensors, cameras, laser scanners, high-powered computers, communications systems ("talking" to other cars) and sophisticated software. Sure, the price of electronics is always dropping, but we're talking about a lot of electronics. How reliable will it be? Will this add a few thousand dollars to the cost of a car, or a few ten thousands of dollars? I'm thinking the early models will be 100K and up. How fast will prices drop?

There are some scanarios which could lead to a faster transition. For example if self-driving cars are inexpensive and gain the public's confidence, people could scrap out conventional cars faster than normal. But let's see when the first cars actually go on sale, and how much they cost before speculating about a transformation of society.

At 11:05 PM, August 18, 2015, Blogger Tory Gattis said...

What could accelerate things, dramatically, would be driverless taxi services getting very big, very quickly - or if transit agencies switched over to on-demand driverless vehicles (which is part of why rail transit investments now don't make sense). People might stop buying new cars or sell their existing cars. It might actually lead to the rapid decline of used car prices if it becomes a market of all sellers and few buyers, and the result of that might be a massive exporting of used cars to other countries. But I agree it's all speculative and could take time.

At 1:40 AM, August 19, 2015, Blogger Kyle said...


Autonomy adds about $18,000 to a car's price in 2015. Add subsidies, cost sharing, improvements, and economies of scale and it probably only bumps up costs negligbly compared to the economic benefits.

At 6:39 AM, August 19, 2015, Blogger Unknown said...

How would autonomous cars deal with road hazards, emergency vehicles, temporary road closures, pedestrians, bicycles, animals, weather, etc? The simple problem is hard to solve, and then the complexity rises. Throw in liability on top of the complexities. It will be interesting to see this develop.

At 10:22 AM, August 19, 2015, Anonymous Mike said...

"Mass adoption of shared rides would solve our traffic congestion problems overnight."

This may be true, but have Americans ever been known to share anything on a mass scale? The vast majority of our country equates being asked to share with communism. How many people have given up their bikes for bike-sharing programs? How many people borrow books from the library instead of buying them? The humorous image of an Uber vehicle with 5-6 separate compartments because Americans can't stand shared space, while true, is case in point.

"So what should cities be doing now to prepare for this future? ... Stop investing in new rail transit."

Oh, this is just a little too convenient, given your known political positions. Trust us folks, if you just stop spending on rail now, this new technology is sure to take care of everything! Why not just tell people to stop buying new cars? Presumably all this mass ride-sharing will eliminate the need for those, won't it?

At 1:33 PM, August 19, 2015, Blogger Tory Gattis said...

Actually, you raise a good point: a new car is an investment in a depreciating asset. If new technology coming down the pike might accelerate that depreciation, it's worth considering before making the leap and buying one. That new car may end up become much more expensive than you expected, once you account for higher annual depreciation.

At 12:41 PM, August 20, 2015, Anonymous Derek said...

This hype over driverless cars is overblown. At best I see it as replacing taxi and uber drivers.

Do you really think commuters are going to want to deal with 1) waiting for a car, 2) surge pricing (unpredictability), and 3) relatively high base fares, when they could just rely on more predictable public transit, manual driving, or bike/walk commute? It's the same reason you don't see people Ubering or taxiing to work every day. Prices would have to be very low (like sub-$1 per trip) and predictive scheduling extremely good to make it worthwhile.

Also, automobiles are the least efficient way to move people, especially if (as Tory pointed out) they're making all kinds of round-trips to take people places. If we want to seriously tackle climate change and fossil fuel dependency, cars can't as cheaply/easily be part of that equation. Self-driving buses and trains are a different story.

At 2:05 PM, August 20, 2015, Blogger Tory Gattis said...

A shared ride would be cheaper - by quite a bit - than driving your own car (imagine if the daily cost of your commute was divided among 3 or more people?). And even if still own your own vehicle, it could be driverless - I think everybody would like to be more productive during their commute time. The cars can also be electric instead of gas - Tesla has already said they're coming.


Post a Comment

<< Home