Policy Implications of Autonomous Vehicles
Policy Implications of Autonomous Vehicles
Policy Implications of Autonomous Vehicles
EX EC U T I V E S UMMARY
Randal OToole is a senior fellow with the Cato Institute and author of Gridlock : Why Were Stuck in Traffic and What to Do about It (Cato Institute,
2010).
If people can
work, read,
surf the web,
or watch
movies or
television
while
traveling, they
may be willing
to spend more
time in
motion, which
in turn could
lead to
another wave
of suburbanization.
INTRODUCTION
Self-driving cars are coming. Cars that can
take over some driving functions, such as steering on highway lanes or controlling speeds to
avoid accidents, are already on the market.
Google has a prototype car that doesnt even
have the option of being operated by humans.
Just as the mass-produced automobile
transformed the 20th century, the autonomous car will transform the 21st. Housing,
work, shopping, and recreation will all dramatically change as mobility no longer depends on
fallible and easily distracted human drivers.
This mobility will contribute to a new birth
of freedom in the United States and around
the world. As Volvo researcher Anders TylmanMikiewicz observes, The core concept of individual mobility is ultimately about freedom.1
One obvious improvement is safety. Highway fatalities have declined by 25 percent since
2005, thanks in part to smart auto technologies such as electronic stability control and antilock brakes.2 The decline in fatalities should
accelerate as motor vehicles become more autonomous.
Other changes will be more subtle. One is
the amount of time people will allot for travel.
One study found that, throughout the world and
throughout history, people have spent an average
of about 1.1 hours per day traveling.3 But if people
can work, read, surf the web, or watch movies or
television while traveling, they may be willing to
spend more time in motion. This could lead to
another wave of suburbanization or exurbanization of both people and jobs.
Partly because the U.S. Department of
Transportation cancelled its research programs on self-driving cars in 1998, most of the
work on autonomous vehicles is being done by
private auto manufacturers, engineering, and
software companies. This is a good thing, as
it has resulted in a variety of competing technologies and innovations.
However, most federal, state, and local
policymakers are acting as if this revolution
isnt happening. For example, not a single
long-range metropolitan transportation plan
DEFINING AUTONOMOUS
VEHICLES
Autonomous vehicles, also known as smart
cars or robocars, use a variety of sensors, computer processors, and databases such as maps
to take over some or all of the functions of
driving from human operators. The National
Highway Traffic Safety Administration and
Society of Automotive Engineers have each
defined several degrees or levels of autonomy;
confusingly, one has five levels while the other
has six, so level 4 under one system is level 5
under the other. In both systems, a level 0 vehicle has no automation while the top level can
operate full time without a human driver.4
Three levels of autonomy are important for
policymakers:
Partial autonomy, in which the car can
take over some of the functions of driving while still requiring the driver to remain fully alert to driving conditions;
High autonomy, in which the car can
drive itself in most conditions but still
may need to call upon a human driver
to deal with extraordinary or unanticipated situations; and
Full autonomy, in which human operation is not needed and may not even be
possible.
Partially autonomous vehicles rely on radar, lasers, infrared, and/or optical sensors
AUTONOMOUS VEHICLE
TIMETABLE
In 2013 Nissan announced that it would
market multiple affordable highly autonomous vehicles by 2020.6 More recently, the
company said it would make available more
advanced partially autonomous cars that can
deal with traffic and change lanes by 2018.7
Google, which has operated its highly autonomous vehicles more than 700,000 miles
since 2009, is even more ambitious, saying it
expects such cars to hit the market as soon as
2017.8 Although Google is building 100 prototypes of a fully autonomous car without any
driving controls, it doesnt expect to manufacture cars itself but hopes to supply software to
automakers.
A survey of more than 200 experts who
attended an automated vehicles symposium
near San Francisco in July 2014 found that the
median date they expected highly automated
vehicles to reach the market was 2020, while
most expected fully automated vehicles to be
available by 2030.9
Of course, reaching the market and dominating the roads are two different things. The
average car on the road is more than ten years
old.10 This suggests that, even if all cars made
after 2020 were highly autonomous, it could
take close to a decade before half the cars on
the road were capable of driving themselves.
But not all cars made in 2020 will be highly
autonomous. Although the costs of manufacturing the hardware required to make new
cars autonomous, including sensors, processors, and other electronics, are expected to
rapidly decline, IHS Automotive predicts that
highly autonomous cars will initially cost at
least $7,000 to $10,000 more than cars without this capability.11 Moreover, to maximize
profits, auto manufacturers tend to introduce
new features on high-end cars first, and slowly
work down the price chain. Thus, the first Nissan-made car to be highly autonomous is more
likely to be a top-of-the-line Infiniti than an
affordable Versa.
Pessimists note that 17 years passed before
the introduction of antilock brake systems and
their use on just 5 percent of cars on the road.
Similarly, 15 years passed before the introduction of adaptive cruise control and 5 percent
market penetration. In this view, it will be
2060 or so before enough cars on the road are
highly or fully autonomous to make a difference in transportation habits and conditions.
However, highly autonomous vehicles offer far more benefits than antilock brakes. After all, surveys show that the vast majority of
American drivers consider themselves to be
above-average drivers and probably consider
Partially
autonomous
vehicles
capable of
accelerating,
braking, and
even steering
are already on
the market,
with some
being
available
for under
$26,000.
Competition
from
aftermarket
manufacturers
will put
pressure on
automakers
to sell
autonomous
cars at
affordable
prices.
themselves immune enough from auto accidents that antilock brakes were only important for other drivers.12
A better comparison is the introduction of
the moving assembly line, which allowed Ford
to sell cars at far lower prices than ever before.
In 1913, when Henry Ford began experimenting with the moving assembly line, less than 5
percent of American families owned an automobile. Fourteen years later, 55 percent owned
a car.13 Wed rather do without clothes than
give up the car, a small-town resident told social scientists in the mid-1920s.14
While the key to growing automobility in
the early 20th century was affordability, there
are several reasons to think that highly autonomous cars will quickly become affordable.
First, software makers such as Google will
have an incentive to prod automakers to make
highly autonomous cars widely available so
they can sell their software, which will likely
be sold on a subscription basis due to the need
to frequently update maps and other data.
Pressure will also come from companies offering aftermarket modifications of existing
cars to make them highly autonomous. While
the cost of converting an older car to a highly
autonomous vehicle would be prohibitive,
turning a partially autonomous car into a highly
autonomous one would require little more than
the addition of a few sensors and a software upgrade. One startup company has already promised to convert partially autonomous Audis to
highly autonomous cars starting in 2015.15
This suggests that, when the first highly
autonomous vehicle hits the market, millions
of cars already on the road will be ready for a
low-cost upgrade to highly autonomous capabilities. This will put pressure on automakers to
sell highly autonomous cars at affordable prices
or lose sales to aftermarket manufacturers.
Furthermore, when fully autonomous vehicles reach the market in, or likely before,
2030, converting highly autonomous vehicles
into fully autonomous ones would take nothing more than a software upgrade. This means
that Americas auto fleet could rapidly become
fully autonomous in just a few years after 2030,
both because of retrofitting and because manufacturers would quickly begin making affordable autonomous cars.
IMPLICATIONS FOR
TRANSPORTATION
The introduction and growing use of partially, highly, and fully autonomous vehicles
will each have significant effects on transportation. Partially autonomous vehicles will help
relieve traffic congestion. Highly autonomous
vehicles will change the way people think
about driving and the costs of travel. Fully autonomous vehicles will allow people who lack
drivers licenses to become as mobile as those
who can drive. Each level will also progressively improve roadway safety.
Adaptive cruise control, an important component of most advanced driver assistance systems, helps relieve the driver of the irksome
task of maintaining a safe distance behind
other vehicles in variable speed or stop-and-go
traffic. But it also provides a useful social benefit in helping to relieve congestion in the first
place. In heavy traffic, much congestion results from slow human reflexes. By substituting computer speeds for human speeds, much
of this congestion may go away.
Actual measurements have found that a
typical freeway lane can move about 2,000 to
2,400 vehicles per hour when moving about
50 miles per hour. However, at slower speeds,
flows can decline to less than half of the full
capacity of the road. This means that if one
vehicle on a road that is near capacity slows
slightly, the entire flow rate on the road can
drop, and speeds can remain slow for hours after the initial slowdown.
Traffic engineers call the transmission of these
slower speeds through a traffic lane a pulse.
Due to its faster response time, a car with adaptive cruise control can interrupt this pulse and
permit flows to remain near capacity levels. One
study has concluded that, if as few as 20 percent
of cars on the road use adaptive cruise control,
about half of all congestion could be avoided.16 A
study in Germany confirms that a small amount
IMPLICATIONS FOR
INFRASTRUCTURE
Telecommunications engineers refer to
infrastructure as being either smart or dumb.
Dumb infrastructure would do little but offer
The
introduction
of fully
autonomous
vehicles will
lead to a
significant
number of
zerooccupant
vehicles on
the road as
people use
cars as robotic
assistants for
a variety of
tasks.
Upgrading
traffic signals
is the most
cost-effective
way of
relieving
congestion,
saving energy,
and
reducing
pollution,
yet few local
governments
have made an
effort to do
so.
Vehicle-to-infrastructure communications
would be a form of smart infrastructure that
could allow roadway owners to communicate
speed limits, traffic signals, and other information, allowing designers of autonomous vehicles to rely on such radio signals rather than
having to program their vehicles to read and
interpret signs and signals of various shapes
and designs. For this to work, however, all state
and local road departments would have to install and maintain such radio systems, something for which there is currently no funding.
Rather than rely on state and local roadway
owners to install such systems, autonomous
vehicle designers say they will use vehicle-toinfrastructure systems if they are available but
wont design their vehicles to depend on them.
The need for mandatory V2V communications in increasing road capacities is questionable. Although freeway lanes have been
measured moving 2,200 vehicles per hour or
more, this much traffic is prone to breakdown
if someone slows down slightly and slow human reflexes force an entire column of traffic
to slow to stop-and-go speeds. For this reason,
managers of variable-priced toll lanes try to
keep traffic below 1,800 vehicles per hour, or
one vehicle every two seconds.
Todays adaptive cruise control can reduce
the gaps between vehicles to 1 second, effectively doubling roadway capacities.27 Advocates of vehicle-to-vehicle communications
say that capacities can be increased further
through V2V communications, known as cooperative adaptive cruise control, which will
allow vehicles to operate in lock-step with one
another in long platoons.
However, mandatory V2V communications systems are not needed for this to happen. First of all, radar-based adaptive cruise
control can respond nearly as fast as cooperative systems. Second, people who wish to
join platoons of close-formation vehicles can
do so by installing optional vehicle-to-vehicle
communication systems. Such systems could
be installed simply by downloading an app to
a smart phone and plugging the smart phone
into the vehicle.
If V2V communications
are required
in new cars, a
decade will
pass before
half the auto
fleet has such
systems,
whereas half
of adults
already have
voluntary
V2V systems
in the form
of smart
phones.
When shared
use of
autonomous
vehicles
becomes
popular, many
people who
take transit
today will
find that
car sharing
makes more
sense.
Although
self-driving
cars may
dominate the
roads in 20
years, not a
single
long-range
regional
transportation
plan has
considered
the effects of
such vehicles
on urban
areas.
10
Rather than
deal with the
uncertain
future, the
current
regional
planning fad is
to design for
the past and
then attempt
to impose that
past using
regulations
and
subsidies.
Yet planners use these models to make longterm, irreversible decisions about urban areas.
Autonomous vehicles introduce the ultimate in uncertainties about future transportation. While planners may not know if gas
prices are going up, they can at least predict
that, if they do go up, the growth in driving
will decline. But such predictions about the
effects of autonomous vehicles are impossible.
No one today has a highly or fully autonomous
vehicle shaping their day-to-day travel, so regional planners cant use travel surveys to answer such questions as
Will autonomous vehicles lead to more
miles of vehicle travel or less?
Will they reduce congestion, and if so
will that make inner cities more attractive because they are less congested? Or
will they lead more people to become
exurbanites?
How will autonomous trucks and increased employee mobility determine
the locations of new job centers?
Will car sharing reduce the demand for
parking?
Will fully autonomous vehicles increase
road capacities enough to reduce the
need for more roads, or will increased
driving offset this benefit?
How will the use of fully autonomous
cars as robotic assistants reshape retail
and service sectors?
How will autonomous vehicles affect
transit?
Planners have techniques to deal with uncertainty, but they arent very effective. One
is to do sensitivity runs of their computer
models with a range of assumptions built in. In
the best case, they hope to find that the results
are not very sensitive to changing the assumptions. If instead the results prove highly sensitive to the assumptions, the best they can do is
guess which assumption is correct.
Rather than deal with the uncertain future,
the current regional planning fad is in effect
to design for the past and then attempt to im-
base because new technologies allow productive utilization of a steadily increasing share
of the trees that are cut. As a result, per capita
demand for timber is declining.
The nation is hardly short of open space.
Federal, state, and local governments own
more than 850 million acres of land, or 38 percent of the total.45 More than 80 percent of
this land is available for recreation and public
enjoyment. Hundreds of millions of acres of
private rural land also provide scenic vistas
and, in many cases, recreation.
In short, urban and even exurban development is no threat to the nations farms, forests,
or open space. The use of urban-growth boundaries and other land-use restrictions to contain
sprawl is needless and costly: nearly all urban
areas that have used them have seen median
housing prices rise by $50,000 to $500,000
or more.46 Since exurbanization is no threat to
farms, forests, or open space, there is little reason to heed calls for more restrictions on urban
development.
CONCLUSION
Autonomous vehicles will transform America. In view of the unpredictable but large
changes ahead, legislators and other policymakers should change the way they view transportation. At the very least, the following recommendations make sense.
First, Congress should eliminate the New
Starts program that gives cities incentives to
build obsolete rail transit systems that will have
no place in a world of car sharing of smart cars on
dumb infrastructure. Rail transit is extraordinarily expensive, doesnt significantly increase transit
riders in most cities where it is built, and will play
an ever-diminishing role in 21st-century cities.47
Second, Congress should also eliminate the
requirement that states and metropolitan planning organizations prepare endless rounds of
long-range regional transportation plans. Planners are no better than anyone else at predicting the future, yet their plans often lock regions
in to irreversible commitments of resources
whose value proves to be far less than promised.
11
Urban and
exurban
development
is no threat to
the nations
farms, forests,
or open space,
and land-use
restrictions
to contain
sprawl are
needless and
costly.
12
2. Early Estimate of Motor Vehicle Traffic Fatalities in 2013, National Highway Traffic Safety Administration, Washington, 2014, p. 1, tinyurl.com/
q83cnnb; Paul A. Eisenstein, Electronic Stability
Control Systems Saving Thousands of Lives, The
Detroit Bureau, November 30, 2012, tinyurl.com/
qchnn9s.
NOTES
3. Andreas Schafer and David G. Victor, The Future Mobility of the World Population, Transportation Research Part A 34, no. 3 (April 2000): 171205.
4. Preliminary Statement of Policy Concerning
Automated Vehicles, National Highway Traffic Safety Administration, Washington, 2013, pp.
45; Bryant Walker-Smith, SAE Levels of Driving Automation, Center for Internet and Society,
December 18, 2013, tinyurl.com/plp38ss.
5. As near as I can tell, the least expensive car
available with these features is a 2015 Subaru Legacy that sells for about $25,800 including delivery
charges.
6. Nissan Announces Unprecedented Autonomous Drive Benchmarks, NissanNews.com,
August 27, 2013, tinyurl.com/kblym6n; Joseph B.
White, Nissan Expects to Market Self-Driving
Cars by 2020, Wall Street Journal, August 27, 2013,
tinyurl.com/le44frl.
7. Carlos Ghosn Outlines Launch Timetable for
Autonomous Drive Technologies, NissanNews.
com, July 16, 2014, tinyurl.com/nsy7czh.
8. Furhaad Shah, Googles Self-Driving Cars Expected in 2017, dataconomy.com, May 28, 2014,
tinyurl.com/pzf3agj.
9. Steven E. Underwood, Automated Vehicles
Forecast, University of Michigan, 2014, pp. 24,
26, 31, tinyurl.com/q8jhrva.
13
Afford a Self-Driving Car? Fast Company, January
31, 2014, tinyurl.com/q5xfldq.
13. David Lux, The History of American Technology: The Automobile Industry, 19201929,
Bryant College, 1998, tinyurl.com/kcex4td.
14. Robert S. Lynd and Helen Merrell Lynd, Middletown: A Study in Contemporary American Culture
(New York: Harcourt, Brace, 1929), p. 255.
15. Liz Gannes, Tiny Startup Cruise Beats
Google to Offer Self-Driving Car Tech to Consumers, re/code, June 23, 2014, tinyurl.com/ktrestv.
16. L. C. Davis, Effect of Adaptive Cruise Control Systems on Traffic Flow, Physical Review 69
(2004): 66110.
17. Arne Kesting et al., Jam-Avoiding Adaptive
Cruise Control (ACC) and Its Impact on Traffic
Dynamics, Traffic and Granular Flow 2005 (Berlin:
Springer, 2007), p. 663.
18. Patricia Mokhtarian and Cynthia Chen, TTB
or Not TTB, That Is the Question: A Review and
Analysis of the Empirical Literature on Travel
Time, Transportation Research Part A 38, nos.
9-10,(NovemberDecember 2004), pp. 64375.
19. Tylman-Mikiewicz.
20. 2012 American Community Survey, Census Bureau, table B08134 for the nation. To calculate
average travel times, I assumed the category less
than 10 minutes averaged 7 minutes while 60 or
more minutes averaged 67 minutes.
21. Rmy Prudhomme and Chang-Woon Lee,
Size, Sprawl, Speed, and the Efficiency of Cities, Urban Studies 36, no. 11 (1999): 184958; Robert Cervero, Efficient Urbanization: Economic
Performance and the Shape of the Metropolis,
Urban Studies 38, no. 10 (2001): 165171.
14
munity Survey, table B08301.
33. 2012 American Community Survey, table B08141.
37. David Brownstone, Key Relationships between the Built Environment and VMT, Transportation Research Board (2008), p. 7, tinyurl.
com/y9mro58.
38. Households by Type, 1940 to Present, Census Bureau, 2013, tinyurl.com/mkoztts.
39. Percent Urban and Rural by State and County
in 2010, Census Bureau, 2012, tinyurl.com/cn7st9j.
40. Cynthia Nickerson et al., Major Uses of
Land in the United States, 2007 (Washington:
Department of Agriculture, 2011), p. 29.
41. Ibid., Table 1.
750.
The Worst of Both: The Rise of High-Cost, Low-Capacity Rail Transit by Randal
OToole (June 3, 2014)
749.
744.
How States Talk Back to Washington and Strengthen American Federalism by John
Dinan (December 3, 2013)
Published by the Cato Institute, Policy Analysis is a regular series evaluating government policies and offering proposals for reform.
Nothing in Policy Analysis should be construed as necessarily reflecting the views of the Cato Institute or as an attempt to aid or hinder
the passage of any bill before Congress. Contact the Cato Institute for reprint permission. All policy studies can be viewed online at
www.cato.org. Additional printed copies of Cato Institute Policy Analysis are $6.00 each ($3.00 each for five or more). To order, call
toll free (800) 767-1241 or email catostore@cato.org.