1 PREPARING A NATION FOR AUTONOMOUS VEHICLES:

2 OPPORTUNITIES, BARRIERS AND POLICY RECOMMENDATIONS FOR

3 CAPITALIZING ON SELF-DRIVEN VEHICLES
4
5 Daniel J. Fagnant
6 The University of Texas at Austin
7 [email protected]
8 512-232-4252
9
10 Kara Kockelman
11 (Corresponding Author)
12 University of Texas at Austin
13 Department of Civil, Architectural and Environmental Engineering
14 301 E. Dean Keeton St
15 Austin, TX 78712
16 [email protected]
17 512-471-0210
18
19 The following paper is a pre-print, the final publication can be
20 found in Transportation Research Part A 77: 167-181, 2015.
21
22
23
24 Key Words: Vehicle automation, autonomous vehicles, cost-benefit analysis, safety,
25 congestion, market penetration, licensing, liability, privacy, cyber security
26
27 Abstract
28
29 Autonomous vehicles (AVs) represent a potentially disruptive and beneficial change to the way
30 in which we travel. This new technology has the potential to impact personal travel across a
31 wide array of impacts including safety, congestion, and travel behavior. All told, major social
32 AV impacts in the form of crash savings, travel time reduction, fuel efficiency and parking
33 benefits are likely on the order of $2,000 per year per AV, or $3,000 eventually increasing to
34 nearly $5,000 when comprehensive crash costs are accounted for.
35
36 Yet barriers to implementation and mass-market penetration remain. Initial costs will likely be
37 unaffordable and licensing and testing standards in the U.S. are being developed at the state
38 level, rather than adopting a national framework, which may lead to inconsistencies across states.
39 Liability regimes remain undefined, security concerns linger, and absent new privacy standards,
40 a default lack of privacy for personal travel may become the norm. Finally, with the advent of
41 this new technology, many impacts, interactions with other components of the transportation
42 system, and implementation details remain uncertain. To address these concerns, research in
43 these areas should be expanded, and the U.S. and other countries should create nationally
44 recognized licensing structures for AVs, and determine appropriate standards for liability,
45 security, and data privacy.
46

1
1 INTRODUCTION
2 Over the past years the automobile and technology industries have made significant leaps in
3 bringing computerization into what has for over a century been exclusively a human function:
4 driving. New cars increasingly include features such as adaptive cruise control and parking assist
5 systems that allow cars to steer themselves into parking spaces. Some companies have pushed
6 the envelope even further by creating almost fully autonomous vehicles (AVs) that can navigate
7 highways and urban environments with almost no direct human input. Assuming that these
8 technologies become successful and enter the mass market, AVs have the potential to
9 dramatically change transportation. This paper serves as an introduction for transportation
10 professionals and policymakers to AV technology, potential impacts, and hurdles.
11
12 AVs may fundamentally alter transport systems. They have the potential to avert deadly crashes,
13 provide mobility to the elderly and disabled, increase road capacity, save fuel, and lower harmful
14 emissions. Complementary trends (in shared rides and vehicles) may lead us from vehicles as an
15 owned product to an on-demand service. Infrastructure investments and operational
16 improvements, travel choices, parking needs, land use patterns, and trucking and other activities
17 will be impacted.
18
19 The proliferation of AVs is far from guaranteed. In addition to technological challenges, other
20 hurdles remain. High costs hamper large-scale production and mass consumer availability, and
21 complex questions remain relating to legal, liability, privacy, licensing, security, and insurance
22 regulation (e.g. KPMG & CAR 2012, ETQ 2012, Grau 2012). While individual U.S. states have
23 been advancing AV legislation through incremental measures (CIS 2012), the federal
24 government’s recent focus has been on setting automation standards for single and combined
25 automation function applications, rather vehicles with limited or fully automated control.
26
27 At the 2012 signing of California’s law enabling AV licensure (SB 1298), Google founder
28 Sergey Brin predicted that Americans could experience AVs within five years (O’Brien 2012).
29 Assuming an additional five years for prices to drop for some degree of mass-market penetration,
30 AVs may be commonplace by 2022, 18 years after the first successful tests. Whether or not
31 consumer adoption comes this fast, policymakers still need to begin to address the multiple
32 unprecedented issues that AVs could bring.
33
34 AVs Today
35 In 2004 DARPA’s (Defense Advanced Research Projects Agency) Grand Challenge’s was
36 launched with the goal of demonstrated AV technical feasibility by navigating a 150-mile route.
37 While the best team completed just over 7 miles, one year later five driverless cars successfully
38 navigated the route. In 2007 six teams finished the new Urban Challenge, with AVs required to
39 obey traffic rules, deal with blocked routes and fixed and moving obstacles, together which
40 provided realistic, every-day-driving scenarios (DARPA 2012). As of April 2013 Google’s self-
41 driving cars have driven over 435,000 miles on California public roads, and numerous
42 manufacturers – including Audi, BMW, Cadillac, Ford, GM, Mercedes-Benz, Nissan, Toyota,
43 Volkswagen, and Volvo – have begun testing driverless systems. Some features necessary for
44 full vehicle automation are now commercially available, including adaptive cruise control
45 (ACC), lane departure warnings, collision avoidance, parking assist systems, and on-board
46 navigation. Europe’s CityMobile2 project is currently demonstrating low-speed AV transit

2
1 applications in dedicated areas in five cities. Additionally, AVs are becoming increasingly
2 common in other sectors, with military, mining and agricultural applications (ETQ 2012), all of
3 which demonstrate the potential for AVs in “traditional” roadway environments.
4
5 Paper Organization
6
7 AVs should have substantial impacts on the transportation system, including benefits to those not
8 directly using AVs. However, barriers remain to well-managed large-scale AV market
9 penetration. This research illuminates these barriers and suggests federal-level policy
10 recommendations for an intelligently planned transition as AVs become a growing share of our
11 transportation system.
12
13 As such, this paper contains three major sections:
14 • AVs’ Potential Benefits,
15 • Barriers to Implementation, and
16 • Policy Recommendations.
17
18 The first section reviews existing literature to ascertain system benefits and impacts, in regards to
19 traffic safety, congestion, and travel behaviors. The information is used to estimate and monetize
20 traveler benefits in the form of crash and congestion reduction as well as parking savings across
21 multiple levels of market penetration. The analysis reflects not only autonomous capabilities for
22 individual vehicles, but also increasingly connected and cooperative vehicles and infrastructure
23 systems. The paper’s second section investigates barriers to adoption, largely from a consumer
24 and regulatory standpoint, rather than technical feasibility. The final section proposes policy
25 recommendations to directly address potential barriers identified earlier.
26
27 AVs’ POTENTIAL BENEFITS
28 AV operations are inherently different from human-driven vehicles. They may be programmed
29 to not break traffic laws. They do not drink and drive. Their reaction times are quicker and they
30 can be optimized to smooth traffic flows, improve fuel economy, and reduce emissions. They
31 can deliver freight and unlicensed travelers to their destinations. This section examines some of
32 the largest potential benefits identified in existing research. The exact extent of these benefits
33 remains unknown, but this paper attempts to estimate these benefits to gauge their magnitude,
34 under varying market penetration levels.
35
36 Safety
37 Autonomous vehicles have the potential to dramatically reduce crashes. More than 40% of fatal
38 crashes involve alcohol, distraction, drug involvement and/or fatigue (NHTSA 2012). Self-
39 driven vehicles should not fall prey to human failings, suggesting the potential for at least a 40%
40 fatal crash rate reduction, everything else constant (such as the levels of long-distance, night-
41 time and poor-weather driving). Such reductions do not reflect crashes due to speeding,
42 aggressive driving, over-compensation, inexperience, slow reaction times, inattention and
43 various other driver shortcomings. Driver error is believed to be the main reason behind over 90
44 percent of all crashes (NHTSA, 2008). Even when the vehicle or roadway environment is the
45 critical reason behind a crash, human factors such as inattention, distraction, or speeding
46 regularly contribute to the crash occurrence and/or injury severity.

3
1
2 The scope of potential benefits is substantial. Over 30 thousand persons die each year in the U.S.
3 in automotive collisions (NHTSA 2012), with 2.2 million crashes resulting in injury (NHTSA
4 2013). At $300 billion, the U.S. annual economic costs of crashes are three times higher than
5 those of congestion (Cambridge Systematics 2011), and safety is highlighted as the #1 goal for
6 transportation in Moving Ahead for Progress in the 21st Century (MAP-21).
7
8 While many driving situations are relatively easy for an autonomous vehicle to handle, designing
9 a system that can perform safely in nearly every situation has been very challenging (e.g.
10 Campbell et al. 2010). For example, recognition of humans in the roadway and object materials
11 is both critical and more difficult for AVs than human drivers (e.g., ETQ 2012, and Farhadi et
12 al., 2009). A person in a roadway may be many sizes, in different positions, and/or partly
13 obscured, complicating sensor recognition. Additionally, evasive decisions should depend on
14 whether an object in the vehicle’s path is a large cardboard box or a large concrete block. When
15 a crash is unavoidable, it is crucial that AVs recognize the objects in their path so they may act
16 accordingly. Liability for these incidents is a major concern and could be a substantial
17 impediment.
18
19 There is also the possibility that some drivers will take their vehicles out of self-driving mode
20 and take control. Ultimately, researchers predict that AVs will overcome many of the obstacles
21 that inhibit them from accurately responding in complex environments. Hayes (2011) suggests
22 that motor-vehicle fatality rates could eventually approach those seen in aviation and rail, about
23 1% of current rates; and KPMG and CAR (2012) advocate a goal of creating “crash-less cars”,
24 while noting that connected vehicle solutions could mitigate up to 80% of unimpaired crashes.
25
26 Congestion and Traffic Operations
27 Aside from safety improvements, researchers are also developing ways for AV technology to
28 reduce congestion and fuel consumption. For example, AVs can sense and possibly anticipate
29 lead vehicles’ braking and acceleration decisions. Such technology allows for smoother braking
30 and fine speed adjustments of following vehicles, leading to fuel savings and reductions in
31 traffic-destabilizing shockwave propagation. AVs may also use existing lanes and intersections
32 more efficiently through shorter headways, coordinated platoons, and more efficient route
33 choices.
34
35 These benefits will not happen automatically. Many congestion-saving improvements depend not
36 only on automated driving capabilities, but also cooperative abilities through vehicle-to-vehicle
37 (V2V) and vehicle-to-infrastructure (V2I) communication. But significant congestion reduction
38 could occur if the safety benefits alone are realized: 25% of congestion is attributable to traffic
39 incidents, around half of which are crashes (FHWA 2005).
40
41 Multiple studies have investigated the potential for AVs to reduce congestion under differing
42 scenarios. Congestion savings due to ACC measures and traffic monitoring systems could
43 potentially smooth traffic flows by seeking to minimize freeway traffic accelerations and
44 braking. This could increase fuel economy and congested traffic speeds respectively by 23% to
45 39% and 8% to 13%, for all vehicles in the freeway travel stream, depending on communication
46 and how traffic smoothing algorithms are implemented (Atiyeh, 2012). If vehicles are enabled to

4
1 travel closer together, the system’s fuel and congestion savings rise further, and some expect a
2 significant increase in highway capacity on existing lanes. Shladover et al. (2012) estimate that
3 cooperative adaptive cruise control (CACC) deployed at a 10%, 50% and 90% market-
4 penetration levels will increase lanes’ effective capacities by around 1%, 21% and 80%,
5 respectively. Headway reductions coupled with near-constant velocities produce more reliable
6 travel times. Similarly, shorter start-up times and headways between vehicles at traffic signals
7 mean that AVs could more effectively utilize green time at signals, improving intersection
8 capacities.
9
10 Of course, many such benefits may not be realized until high AV shares are present. For
11 example, if 10% of all vehicles on a given freeway segment are AVs, there will likely be an AV
12 in every lane at regular spacing during congested times to smooth traffic for all travelers.
13 However, if just one out of two hundred vehicles are AVs, the impact will be greatly lessened.
14 Also, if one AV is following another, the following AV can reduce the headway between the two
15 vehicles, increasing roadway capacity. This efficiency benefit is also contingent upon higher AV
16 shares. Technical and implementation challenges also loom in order to realize the full potential
17 of high adoption shares, including city-wide or regional coordination paradigms.
18
19 Travel-Behavior Impacts
20 Like safety and congestion, travel behavior may also change significantly. AVs may provide
21 mobility for those too young to drive, the elderly and the disabled, thus generating new roadway
22 demands. Parking patterns could change as AVs self-park in less-expensive areas. Car- and ride-
23 sharing programs could expand, as AVs serve multiple persons on demand; and the trucking
24 industry may realize better fuel savings via road-trains, or even one day go driverless. Most of
25 these ideas point toward more vehicle-miles traveled (VMT) and automobile-oriented
26 development, though perhaps with fewer vehicles and parking spaces.
27
28 As of January 2013 in California, Florida and Nevada, legislation mandates that all drivers
29 pursuing AV testing on public roadways be licensed and prepared to take over vehicle operation,
30 if required. As AV experience increases, this requirement could fall away and AVs could be able
31 to legally chauffeur children and persons that otherwise would be unable to safely drive. Such
32 mobility may be increasingly beneficial, as the U.S. population ages, with 40 million Americans
33 over the age of 65 and this demographic growing 50% faster than the overall population (U.S.
34 Census 2011). Wood (2002) observes many drivers cope with physical limitations through self-
35 regulation, avoiding heavy traffic, unfamiliar roads, night-time driving, and poor weather; while
36 others stop driving altogether. AVs could facilitate personal independence and mobility while
37 maintaining safety, thus further increasing automobile travel demand.
38
39 With increased mobility among the elderly and others, as well as lowered congestion delays,
40 VMT increases may be expected along with associated congestion, emissions, and crash rates,
41 unless demand-management strategies are thoughtfully implemented (Kockelman and Kalmanje
42 2006). Most AV benefits will likely exceed the negative impacts of added VMT. For example,
43 even if VMT were to double, a reduction in crash rates per mile traveled by 90% yields a
44 reduction in the total number of crashes and their associated traffic delays by 80%. Likewise,
45 unless new AV travel is truly excessive, highway capacity improvements should accommodate
46 the new/induced demand, thanks to AVs’ congestion-mitigating features (like traffic smoothing

5
1 algorithms) and capacity-increases (through CACC), as well as public-infrastructure investments
2 (like V2I communication systems with traffics signals).
3
4 Already-congested traffic signals and other roadway infrastructure could be negatively impacted
5 due to increased trip-making. However AVs could enable smarter routing in coordination with
6 intelligent infrastructure, quicker reaction times, and closer spacing between vehicles to
7 counteract increased demand. Whether arterial congestion improves or degrades ultimately
8 depends on how much VMT is induced, the magnitude of AV benefits, and demand management
9 strategies like road pricing. Emissions fall when travel is smooth, with Berry (2010) estimating
10 that a 20% reduction in accelerations and decelerations should lead to 5% fuel consumption
11 reductions, and resulting emissions, so while AVs may increase VMT, emissions per mile could
12 fall.
13
14 Additional fuel savings may accrue through smart parking (Bullis 2011). In-vehicle systems
15 could communicate with parking infrastructure and enable driverless drop-offs and pickups. This
16 same technology could improve and expand car sharing and dynamic ride-sharing by allowing
17 for nearby, real-time rentals on a per-minute or per-mile basis. If successful, this business model
18 may explode since users could simply order an on-demand taxi using mobile devices.
19 Preliminary results (Fagnant and Kockelman 2013) using an agent-based model for assigning
20 vehicles around a region in combination with NHTS data (FHWA 2009) indicate that a single
21 shared AV could replace between nine and thirteen vehicles owned by individual households,
22 without comprising current travel patterns. As shown in Figure 1, even in Seattle where vehicle
23 use is more intense than national averages (PSRC 2006, FHWA 2009), just than 11% of vehicles
24 are “in use” throughout the day, even at peak times, though usage rises to 16% if only including
25 newer vehicles.
26
18%
16%
14%
12%
10%
8% Vehs ≤ 10 yrs
6% Vehs ≤ 15 yrs
4%
All vehs
2%
0%
3:00 AM

6:00 AM

9:00 AM

3:00 PM

6:00 PM

9:00 PM
Midnight

Noon

Midnight

27
28 Figure 1: Vehicle Use by Time of Day and by Vehicle Age
29 (Source: PSRC 2006 household travel survey data)
30
31 Freight
32 Freight transport on and off the road will also be impacted. Mining company Rio Tinto already
33 uses ten self-driving ore trucks, with plans to expand to 150 vehicles (ETQ 2012). Technologies
34 that apply to autonomous cars can also apply to the trucking industry, increasing fuel economy
35 and potentially the need for drivers. While workers must still load and unload cargo, long-

6
1 distance driverless journeys may be made, with warehousing employees handling shipments at
2 either end. This is not to claim all trucks will become driverless, but AVs could dramatically
3 change the industry, enhancing productivity and lowering costs through reduced labor and
4 improved service times. Political resistance may rise from labor groups as well as competing
5 industries such as the freight railroads.
6
7 Higher fuel economies may emerge when using tightly coupled road-train platoons, from
8 reduced air resistance. Lowered travel times from higher capacity networks may also be
9 realized, as noted previously. Bullis (2011) estimates that 4-meter inter-truck spacings could
10 reduce fuel consumption by 10 to 15 percent, and road-train platoons facilitate adaptive braking,
11 enabling further fuel savings. Kunze et al. (2009) successfully demonstrated 10-meter headways
12 between multiple trucks on public German motorways, and platooned Volvo trucks recently
13 logged 10,000 km in Span (Newcomb, 2012).
14
15 Anticipating AV Impacts
16 Since AVs are only in the testing phase, it is difficult to precisely anticipate outcomes.
17 Nevertheless, it is useful to roughly estimate likely impact magnitudes. Based on research
18 estimates for potential impacts discussed above, this paper quantifies crash, congestion and other
19 impacts for the U.S. transportation network (including parking, VMT, and vehicle counts). The
20 analysis assumes three AV market penetration shares: 10%, 50% and 90%. These represent not
21 only market shares, but also assume technological improvement over time, as it could take many
22 years to see high penetration rates. This analysis is inherently imprecise, but gives the reader an
23 order of magnitude estimate of AVs’ broad economic and safety impacts.
24
25 VMT Change
26
27 VMT per AV is assumed to be 20% higher than that of non-AV vehicles at the 10% market
28 penetration rate, and 10% higher at the 90% market penetration rate. This reflects the fact that
29 early adopters will generally have more pent-up demand for such vehicles than later buyers.
30 Preliminary simulations (Fagnant and Kockelman 2013) underscore this idea, finding that a fleet
31 of shared AVs serving around 65 thousand trips per day (representing over 2% of regional trips
32 across a simulated city grid) cover 10.2% of their daily travel unoccupied, with this figure falling
33 to 6.6% as the number of trips served doubles (thanks to a higher intensity of nearby pickups and
34 drop-offs).
35
36 Additional VMT increases may be realized from induced demand as travel costs and congestion
37 fall. In his thirty year review of literature, Cervero (2001) shows that the long term elasticity of
38 VMT demand with respect to lane miles ranges from 0.47 to 1.0, averaging 0.74. This means
39 that if regional vehicle lane mile increase by 1%, VMT should increase by around 0.74%. While
40 AVs’ congestion impacts are similar to increased lane miles, the effective capacity expansion is
41 uniform, rather than targeted. That is, many areas today are uncongested, do not have latent
42 demand and will likely continue to be so for the foreseeable future. This report does not account
43 for induced travel due to latent demand, which may be stemmed with policies like congestion
44 pricing. However, if even half of Cervo’s elasticity estimates are applied, system-wide VMT
45 could experience 37% growth by the 90% AV market-penetration level, from increased capacity
46 effects.

7
1
2 Discount Rate and Technology Costs
3
4 For net-present value calculations, a 10% discount rate was assumed, which is higher than the
5 7% rate required by the Office of Management and Budget (OMB) for federal projects and
6 TIGER grant applications (LaHood 2011) in order to reflect the greater uncertainty of this
7 emerging technology. Initial costs at the 10% market penetration level were assumed to add
8 between $10,000 to the purchase price of a new vehicle, falling to $3,000 by the 90% market-
9 penetration share, as noted later in this paper’s Vehicle Cost section.
10
11 Safety
12
13 This analysis assumes 10% of AVs are shared (at all levels of penetration), and that a single
14 shared AV serves ten times as many trips as a non-shared vehicle. U.S. crash rates and severity
15 distributions for non-AVs are assumed constant and unchanged, based on NHTSA’s 2011 values.
16 As noted previously, over 90% of the primary factors behind crashes are due to human errors
17 (NHTSA 2008), and 40% of fatal crashes involve driver alcohol or drug use, driver distraction
18 and/or fatigue (NHTSA 2011). Therefore, AVs may be assumed 50% safer than non-AVs at the
19 early, 10% market penetration rate (reflecting savings due to eliminating these factors, as well as
20 fewer legal violations like running red lights), and 90% safer at the 90% market penetration rate
21 (reflecting the near-elimination of human errors as primary crash causes, greater V2V use and
22 improving technologies). Pedestrians and cyclists are assumed to enjoy half the AVs’ safety
23 benefits, since just the driver relies on the AV technology. Similarly, motorcyclists may be
24 lagging adopters (with technological implementation issues too), and around half of all fatal
25 motorcycle crashes do not involve another vehicle. Therefore, motorcycles are assumed to
26 experience just a 25% decline in their crash rates compared to other vehicles. Crash costs were
27 estimated based on economic consequences (NSC 2012), and also on higher comprehensive
28 costs, as recommended by the U.S. DOT (Trottenberg, 2011), to reflect pain, suffering and the
29 statistical value life.
30
31 Congestion Reduction
32
33 Shrank and Lomax’s (2011) congestion impact projections for 2020 were used as a baseline.
34 They assumed a $17 per hour per traveler value of travel time, $87 per hour of truck travel time,
35 and 2010 statewide average gas prices. They estimate that 40% of the nation’s roadway
36 congestion occurs on freeways (with the remainder on other streets), and that by 2020, U.S.
37 travelers will experience around 8.4 billion hours of delay while wasting 4.5 billion gallons fuel
38 (due to congestion), for annual economic costs of $199 billion.
39
40 Here, it is assumed that AVs are equipped with CACC and traffic flow smoothing capabilities.
41 At the 10% AV-market penetration level, freeway congestion delays for all vehicles are
42 estimated to fall 15%, mostly due to smoothed flow and bottleneck reductions. This is lower
43 than Atiyeh (2012) suggests, to reflect induced travel, though additional congestion benefits may
44 also be realized (due to fewer crashes, slight increased capacity from CACC, and better routing
45 choices). At the 50% market penetration level, a cloud-based system is assumed to be in use
46 (Atiyeh suggests 39% congestion improvements from smoothed flow), and further capacity

8
1 enhancements of 20% may be realized (Shladover et al. 2012). Furthermore, with crashes falling
2 due to safety improvements, another 4.5% congestion reduction may be obtained. Again,
3 induced travel will counteract some of these benefits and a 35% freeway delay reduction is
4 estimated. Finally, at the 90% level, freeway congestion is assumed to be reduced by 60% with
5 the near doubling of roadway capacity and dramatic crash reductions. However, readers should
6 note that capacity and delay are not linearly related, and congestion abatement may be even
7 greater than these predictions with 90% market penetration.
8
9 At the arterial-roadway level, congestion is assumed to benefit much less, since delays emerge
10 largely from conflicting turning movements, pedestrians, and other transportation features that
11 AV technologies cannot address as easily. Therefore, arterial congestion benefits are assumed to
12 be just 5% at the 10% market-penetration level, 10% at the 50% penetration rate, and 15% at
13 90% market penetration. AV fuel efficiency benefits are assumed to begin at 13%, increasing to
14 25% with 90% market penetration, due to better route choices, less congestion, road-train drag
15 reductions (from drafting), and more optimal drive cycles. Non-AVs on freeways are assumed to
16 experience 8% fuel economy benefits during congested times with 10% market penetration, and
17 13% at the 50% and 90% penetration levels. For simplicity, this analysis assumes induced
18 travel’s added fuel consumption is fully offset by AVs’ fuel savings benefits during non-
19 congested times.
20
21 Parking
22
23 Parking savings comprise this analysis’ final monetized component. Litman (2012) estimates
24 that comprehensive (land, construction, maintenance and operation) annual parking costs are
25 roughly $3,300 to $5,600 per parking space in central business districts (CBDs), $1,400 to
26 $3,700 per parking space in other central/urban areas, and $680 to $2,400 per space in suburban
27 locations. Therefore moving a parking space outside of the CBD saves nearly $2,000 in
28 annualized costs, while moving one to a suburban location save another $1,000. In addition to
29 moved spaces, fewer overall spaces should be needed thanks to car sharing. Therefore, while not
30 every AV will result in a moved or eliminated parking space, this analysis assumes that $250 in
31 parking savings will be realized per new AV.
32
33 Summary Economic Analysis
34
35 Table 1 summarizes all of these estimated impacts, suggesting economic benefits reaching $189
36 billion ($434 billion, comprehensive) with a 90% AV market penetration rate. Meaningful
37 congestion benefits are estimated to accrue early on, while crash benefits magnitude grows over
38 time. For example, congestion savings represent 66% of benefits and crash savings represent
39 21% of benefits -- at the 10% market penetration level, versus 33% and 58% of benefits,
40 respectively, at the 90% penetration rate. When including comprehensive costs, crash savings
41 more than triple.
42
43
44
45 Table 1: Estimates of Annual Economic Benefits from AVs in the United States
Assumed Market Shares

9
10% 50% 90%
Crash Cost Savings from AVs
Lives Saved (per year) 1,100 9,600 21,700
Fewer Crashes 211,000 1,880,000 4,220,000
Economic Cost Savings $5.5 B $48.8 B $109.7 B
Comprehensive Cost Savings $17.7 B $158.1 B $355.4 B
Economic Cost Savings per AV $460 $1,080 $1,690
Comprehensive Cost Savings per AV $1,470 $3,500 $5,460
Congestion Costs
Travel Time Savings (M Hours) 756 1680 2772
Fuel Savings (M Gallons) 102 224 724
Total Savings $16.8 $37.4 $63.0
Savings per AV $1,400 $830 $970
Other AV Impacts
Parking Savings $3.0 $11.3 $16.3
Savings per AV $250 $250 $250
VMT Increase 2.0% 7.5% 9.0%
Change in Total # Vehicles -8.3% -31.0% -44.8%
Annual Savings: Economic Costs Only $25.3 B $97.5 B $189.0 B
Annual Savings: Comprehensive Costs $37.6 B $206.8 B $434.7 B
Savings Per AV: Economic Costs Only $2,110 $2,160 $2,910
Savings Per AV: Comprehensive Costs $3,120 $4,580 $6,680
Net Present Value of AV Benefits minus
Purchase Price: Economic Costs Only $6,050 $11,430 $19,130
Net Present Value of AV Benefits minus
Purchase Price: Comprehensive Costs $13,730 $29,840 $47,810

Assumptions
Number of AVs Operating in U.S. 0.5 0.75 0.9
Crash Reduction Fraction per AV 15% 35% 60%
Freeway Congestion Benefit (delay reduction) 5% 10% 15%
Arterial Congestion Benefit 13% 18% 25%
Fuel Efficiency Benefit 8% 13% 13%
Non-AV Following-Vehicle Fuel Efficiency
Benefit (Freeway) 20% 15% 10%
VMT Increase per AV 10% 10% 10%
% of AVs Shared across Users 10% 10% 10%
Added Purchase Price for AV Capabilities $10,000 $5,000 $3,000
Discount Rate 10% 10% 10%
Vehicle Lifetime (years) 15 15 15
1
2 Table 1 illuminates AVs’ social benefits, though it is also meaningful to examine privately
3 realized AV benefits. Table 1’s assumptions at the 10% market penetration level may be

10
1 compared to $2,400 annual fuel costs and $1,000 insurance costs (AAA 2012), as well as various
2 potential parking savings over 250 annual work days. Other benefits include valuations for time
3 driven autonomously, with total annual vehicle hours traveled based on U.S. averages (14,200
4 miles per year) divided by an assumed average speed of 30 mph (FHWA 2013). This results in
5 the ranges of benefits shown in Table 2, across various purchase prices, values of time and
6 parking costs:
7
8 Table 2: Privately Realized Internal Rates of Return
Benefits (Daily Parking & Hourly Value of Travel Time)
Added Costs
$0 & $0 $0 & $1 $1 & $1 $5 & $1 $1 & $5 $5 & $5 $5 & $10 $10 & $10

$100k+ -19% -16% -14% -10% -7% -5% 0% 2%

$37.5k -12% -7% -6% 0% 4% 8% 16% 20%
$10k 3% 10% 13% 24% 34% 44% 68% 80%
9
10 At high technology costs of $100,000 or more, benefits are mostly small compared to purchase
11 prices, except for individuals with very high values of time. Once prices come down to $37,000,
12 however, affluent persons with high values of travel time and/or parking costs may find the
13 technology a worthwhile investment. Only at the $10,000 added price does the technology begin
14 to truly become a realistic investment for many, with even a $1 per hour time value and $1 daily
15 parking cost generating a 13% rate of return.
16
17 It should be noted that this report does not quantify or monetize several of the impacts discussed
18 earlier. Benefits to the newly mobile are not forecast, nor health impacts from diminished
19 walking distances. Some of the nation’s 240,000 taxi drivers and 1.6 million truck drivers (BLS
20 2012) could be displaced by AVs, while emissions, infrastructure needs, and walking rates may
21 change depending on induced VMT. Sprawl or automobile-style development could also result,
22 which are not included in the analysis.
23
24 While exact magnitudes of all impacts remain uncertain, this analysis shows the potential for
25 AVs to deliver substantial benefits to many, thanks to sizable safety and congestion savings.
26 Even at 10% market penetration, this technology has the potential to save over a thousand lives
27 per year and offer tens of billions of dollars in economic gains.
28
29 BARRIERS TO IMPLEMENTATION
30 AVs present many opportunities, benefits and challenges while ushering in new behavioral
31 changes. The speed and nature of the transition to a largely AV system are far from guaranteed;
32 they will depend heavily on purchase costs, as well as licensing and liability requirements.
33 Moreover, AVs present security and privacy risks. Even with a smooth and rapid deployment
34 that addresses security and privacy concerns, a system that optimally exploits AV capabilities
35 requires special research efforts. The following outlines these barriers.
36
37 Vehicle Costs
38 One barrier to large-scale market adoption is AV cost. Technology needs include sensors,
39 communication and guidance technology, and software for each automobile. KPMG and CAR

11
1 (2012) note that the Light Detection and Ranging (LIDAR) systems atop Google’s AVs cost
2 $70,000, with further costs from other sensors, software, engineering, and added power and
3 computing requirements. Dellenback (2013) estimates that most current civilian and military AV
4 applications cost over $100,000. This is simply unaffordable for most Americans, with 2012
5 sticker prices for the top 27 selling vehicles in America ranging from $16,000 to $27,000
6 (Boesler 2012).
7
8 As with electric vehicles, technological advances and large-scale production promise greater
9 affordability over time. Dellenback (2013) estimates that added costs may fall to between $25
10 and $50,000 with mass production, and likely will not fall to $10,000 for at least 10 years.
11 Insurance, fuel, and parking-cost savings may cover much of the added investment, as noted
12 earlier. Typical annual ownership and operating costs ranged from $6,000 to $13,000, depending
13 on vehicle model and mileage (AAA 2012).
14
15 If AV prices come close to conventional vehicle prices, research suggests a ready market for
16 AVs. J.D. Power and Associates’ (2012) recent survey suggests that 37% of persons would
17 “definitely” or “probably” purchase a vehicle equipped with autonomous driving capabilities in
18 their next vehicle, though the share dropped to 20% after being asked to assume an additional
19 $3000 purchase price. Volvo senior engineer Erik Coelingh estimates the same $3000 mark for
20 AV capabilities (ETQ 2012), though early adopters will likely pay much more, as noted above.
21 For comparison, as of February 2013, adding all available driver-assist features, adaptive cruise
22 control, safety options (including night vision with pedestrian detection) and the full “technology
23 package” increases a BMW 528i sedan’s purchase price by $12,450, from a base MSRP of
24 $47,800 (BMWUSA 2013). Of course, while these features provide guidance and a degree of
25 automation for certain functions, full control remains with the human driver.
26
27 As AVs migrate from custom retrofits to mass-produced designs, it is possible that these costs
28 could fall somewhere close to Coelingh and J.D. and Associates’ $3,000 mark, or, eventually,
29 perhaps just $1,000 to $1,500 more per vehicle (KPMG and CAR 2012). Nevertheless, cost
30 remains a key implementation challenge, due to the current unaffordability of even some of the
31 more basic technologies.
32
33 AV Licensing
34 As of July 2013, California, Nevada and Washington D.C. have enacted legislation allowing AV
35 licensing, and Florida enables AV testing, with related legislation pending in another ten states
36 (CIS 2013). States have thus far declined to set many specific restrictions, directing their state
37 Departments of Motor Vehicles (DMVs) to establish regulatory licensing and provisional testing
38 standards. This legislative guidance has varied significantly, from state to state. For example,
39 Nevada’s legislation contains just 23 lines of definitions and broad guidance to its DMV, while
40 California’s is a more detailed 6 pages and similar direction to their DMV (to establish safety
41 and testing specifications and requirements). Without a consistent (or at least congruent)
42 licensing framework and safety standardization for acceptance, AV manufacturers may face
43 regulatory uncertainty and unnecessary overlap.
44
45 California’s more detailed legislative content provides concrete requirements for AVs. It
46 includes specific requirements for AV testing on public roads, including insurance bonding, the

12
1 ability to quickly engage manual driving, failsafe systems in case of technology failure, and
2 sensor data storage prior to any collision. The DMV must consider a broad array of regulations,
3 including the number of AVs on California’s public roads, AV registration numbers, AV
4 operator licensing and requirements, possible AV license revocations, and licensing denial.
5 Finally, the legislation requires public hearings and directs the DMV to enact strict AV
6 oversight.
7
8 While California’s DMV rulemaking is expected by 2015, Nevada has already processed AV
9 testing licenses (on public roads) for Google, Continental and Audi. These licensing
10 requirements include a minimum of 10,000 autonomously driven miles and documentation of
11 vehicle operations in a number of complex situations. Furthermore, Nevada can grant testing
12 licenses subject to certain geographic and/or environmental limitations (e.g., autonomous
13 operation only on the state’s interstates, for daytime driving free of snow and ice). While the
14 strategies pursued by these states is groundbreaking, if disparate versions of these regulatory
15 issues emerge (across distinct states), AV manufacturers will incur delays and increased
16 production and testing costs.
17
18 Customarily, drivers licensed in one U.S. state are able to legally operate a vehicle in other states
19 through reciprocity agreements, as outlined in the state Driver License Compact, constituting
20 agreements between all but five U.S. states (Georgia, Wisconsin, Massachusetts, Michigan, and
21 Tennessee). The language states that “It is the policy of each of the party states to… make the
22 reciprocal recognition of licenses to drive… in any of the party states” (State of Montana 2011).
23 Smith (2012, p. 95) argues that current law probably does not prohibit automated vehicles, even
24 in states without explicit licensing, though failure to clarify regulations may “discourage their
25 introduction or complicate their operation.”
26
27 Litigation and Liability
28 A vehicle driven by a computer on public roads opens the possibility of many insurance and
29 liability issues. Even with near-perfect automated driving, there may be instances in which a
30 crash is unavoidable. For example, if a deer jumps in front of the car, does the AV hit the deer or
31 run off the road? How do actions change if the object is another car, a heavy-duty truck, a
32 motorcyclist, bicyclist or pedestrian? Does the roadside environment and/or pavement wetness
33 factor into the decision? What if the lane departure means striking another vehicle? With a split
34 second for decision-making, human drivers typically are not held at fault when responding to
35 circumstances beyond their control, regardless of whether their decision was the best. In contrast,
36 AVs have sensors, visual interpretation software, and algorithms that enable them to potentially
37 make informed decisions. Such decisions may be questioned in court, even if the AV is
38 technically not “at fault”.
39
40 If AVs are held to a much higher standard that human drivers, AV costs will rise and fewer
41 people will be able to purchase them. Some steps have been made to account for liability
42 concerns. California law (CIS 2013) requires 30 seconds of sensor data storage prior to a
43 collision to help establish fault, assuming that the AV has been programmed and tested properly.
44 Related technologies like parking assist and adaptive cruise control may provide test cases to
45 guide how fully autonomous technologies will be held liable.
46

13
1 Security
2 Transportation policymakers, auto manufacturers, and future AV drivers often worry about
3 electronic security. Computer hackers, disgruntled employees, terrorist organizations, and/or
4 hostile nations may target AVs and intelligent transportation systems more generally, causing
5 collisions and traffic disruptions. As one worst-case scenario, a two-stage computer virus could
6 be programmed to first disseminate a dormant program across vehicles over a week-long period,
7 infecting virtually the entire U.S. fleet, and then cause all in-use AVs to simultaneously speed up
8 to 70 mph and veer left. Since each AV in the fleet represents an access point into such systems,
9 it may be infeasible to create a system that is completely secure.
10
11 To understand the extent of this threat, the problem can be viewed from an effort-and-impact
12 perspective, recognizing the mitigation techniques used in comparable critical infrastructure
13 systems of national importance. According to Jason Hickey (2012), vice president of software
14 security firm Vínsula, current cyber-attacks are more commonly acts of espionage (gaining
15 unauthorized system access for information) rather than sabotage (compromising a system’s
16 operation). Disrupting a vehicle’s communication or sensors, for example, would require a more
17 complex and sophisticated attack than simply gathering information, and disrupting the control
18 commands would be harder still. Engineering an attack to simultaneously compromise a fleet of
19 vehicles, whether from a point source (for example, compromising all vehicles near an infected
20 AV) or from a system-wide broadcast over infected infrastructure, would likely pose even
21 greater challenges for a would-be attacker. Regardless, the threat is real and a security breach
22 could have lasting repercussions.
23
24 Fortunately, robust defenses should make attacks difficult to stage. The U.S. has demonstrated
25 that it is possible to maintain and secure large, critical, national infrastructure systems, including
26 power grids and air traffic control systems. The National Institute of Standards and Technology
27 (NIST) is currently developing a framework to improve critical infrastructure cybersecurity, and
28 recommendations that stem from this framework may be incorporated into automated and
29 connected vehicle technologies. While security measures for personal computers and internet
30 communication were implemented largely as an afterthought, and in an ad-hoc manner (Hickey
31 2012), V2V and V2I protocols were developed with security implemented in the initial
32 development phase (NHTSA 2011). These and other security measures (like the separation of
33 mission-critical and communication systems) should make large-scale attacks on AVs and
34 related infrastructure difficult (Grau 2012 and Hickey 2012). Though Grau (2012) and Hickey
35 (2012) acknowledge that there is no “silver bullet”, such measures make attacks more difficult
36 while limiting potential damage.
37
38 Privacy
39 California-based consumer education and advocacy organization Consumer Watchdog raised
40 privacy concerns during a recent round of AV-enabling legislation (Brandon 2012). Such
41 concerns will likely grow, as AVs and non-autonomous connected vehicles become mainstream
42 and data sharing becomes commonplace. Four primary data-related questions arise: what type of
43 data will be shared, with whom will it be shared, in what way will the data be made available,
44 and for what ends will it be used?
45

14
1 Crash data will likely be available to AV technology suppliers, since they will likely be liable in
2 the event of an AV-caused crash. If a human is driving a vehicle with autonomous capabilities
3 when the crash occurs, however, privacy concerns arise. Few people want their own vehicle’s
4 data recorder being used against them in court, though this merely extends an existing issue: 80%
5 of vehicles sold in the U.S. today have similar (but less detailed) event data recorders that
6 describe vehicle actions taken up to five seconds prior to a crash (The Economist 2012).
7
8 Providing AV travel data including routes, destinations, and departure times to centralized and
9 governmentally controlled systems is likely more controversial, particularly if the data is
10 recorded and stored. Without safeguards, this data could be misused by government employees
11 for stalking individuals, or provided to law enforcement agencies for unchecked monitoring and
12 surveillance. Vehicle travel data has wide-ranging commercial applications that may be
13 disconcerting to individuals, like targeted advertising.
14
15 At the same time, responsible dissemination and use of AV data can help transportation network
16 managers and designers. Data help facilitate a shift from a gas-tax to a vehicle-miles traveled
17 (VMT) fee, or potentially implement congestion pricing schemes by location and time of day.
18 Those who program traffic signals, for example, could use such data to improve efficiency and
19 reduce delays. In contrast, continuously connected AVs or connected conventional vehicles
20 could illuminate vehicle paths and speed changes, and inform signal systems in real time.
21 Likewise this data could be used to assist planners evaluating future improvements, facilitating
22 more effective investment decision-making. Law enforcement could also benefit, and
23 commercial advertising profits may drive down prices. Any decisions to enhance traveler privacy
24 ideally should be balanced against the benefits of shared data.
25
26 Missing Research
27 While AVs may be commercially available within five years, related research lags in many
28 regards. Much of this is due to the uncertainty inherent in new contexts: with the exception of a
29 few test vehicles, AVs are not yet present in traffic streams, and it is difficult to reliably predict
30 the future following such disruptive paradigm shifts. Moreover, technical developments, along
31 with relevant policy actions will impact outcomes, creating greater uncertainty. With these
32 caveats in mind, it is useful to identify the critical gaps in existing investigations to better prepare
33 for AVs’ arrival.
34
35 One of the most pressing needs is a comprehensive market penetration evaluation. As KPMG
36 and CAR (2012) make clear, AVs will be driving on our streets within the next decade, but it is
37 uncertain when they will comprise a substantial share of the U.S. fleet. Market penetration
38 estimates could attach dates and percentages to aggressive, likely, and conservative AV-adoption
39 scenarios. This would provide transportation planners and policy-makers with a reasonable range
40 of outcomes for evaluating competing infrastructure investments, AV policies, and other
41 decisions.
42
43 Other important research gaps include predicting how travel demand patterns will change, how
44 intersections can best be managed (as initially examined by Dresner and Stone [2008]), and how
45 VMT and vehicle emissions will change. With all such estimates in hand, regional planners can

15
1 incorporate many AV impacts in their travel demand models, traffic delay forecasts, air quality
2 estimates, and related decision-making processes.
3
4 POLICY RECOMMENDATIONS
5 Given the apparent promise of AVs, policymakers and the public would be wise to seek a
6 smooth and intelligently planned introduction for and transition to this new technology. AV
7 technology seems likely to advance with or without legislative or agency actions. However, the
8 manner in which AV technologies progress and will eventually be implemented depend on these
9 efforts. Intelligent planning, meaningful vision, regulatory action, and reform are required to
10 address the issues identified above. As such, this report recommends three concrete actions:
11
12 1. Expand Autonomous Vehicle Research
13 Car manufacturers have poured resources into AV technology research and development.
14 Meanwhile research into the impacts that these vehicles could deliver to the transportation
15 system is relatively scarce. This paper has identified key missing links in AV research, including:
16
17 • Future AV market penetration rates;
18 • Travel and land use pattern evolution in the face of AV car-sharing and ride sharing
19 options;
20 • Emissions and energy impacts of AV operations; and
21 • Integrated AV and ITS infrastructure investigations, including facilitation of mileage
22 based user fees.
23
24 Other gaps will become apparent in coming year and as AVs enter the marketplace. It becomes
25 imperative that agencies around the world and at the federal, state and local level, as well as
26 other stakeholders help fund such research to enable regions and nations to anticipate, and more
27 effectively plan for AV opportunities and impacts.
28
29 2. Develop Federal Guidelines for Autonomous Vehicle Licensing
30 To facilitate regulatory consistency, the U.S. DOT should develop a framework and set of
31 national guidelines for AV licensing at the state level. With a more uniform set of standards in
32 place, states can pool efforts developing safety, operational, and other requirements. AV
33 manufacturers will be better able to meet detailed national requirements, rather than matching 50
34 potentially different certification regimes across states. Existing state licensing can provide
35 guidance for such efforts, which will streamline AV licensing and testing. Moreover, AV
36 licensing consistencies could help limit AV product liability, as argued by Kalra et al. (2009).
37
38 3. Determine Appropriate Standards for Liability, Security, and Data Privacy
39 Liability, security, and privacy concerns represent substantial barriers to widespread AV
40 technology implementation. These issues should be addressed to give manufactures and investors
41 more certainty in development. Liability standards should strike the balance between assigning
42 responsibility to manufacturers without putting undue pressure on their product. Robust cyber
43 security standards will help the industry develop ways to prevent outside attacks.
44
45 AV technology consumers will likely have concerns about use and potential abuse of data
46 collected from their personal travel. Therefore, AV-enabling legislation should balance

16
1 legitimate privacy concerns against potential data use benefits. Since vehicles will inevitably
2 cross state boundaries, federal regulation should establish parameters for what types of AV data
3 to share, with whom it should be shared, how the data will be made available, and for what ends
4 it may be used -- rather than take a default (no action) position, which will likely result in few to
5 no privacy protections.
6
7 CONCLUSIONS
8 Driverless cars may seem a distant possibility. In reality, autonomous technology is improving
9 quickly, as some automated features are already on current models. This new technology should
10 reduce crashes, ease congestion, improve fuel economy, reduce parking needs, bring mobility to
11 those unable to drive, and eventually revolutionize travel. Based on current research, annual U.S.
12 economic benefits could be around $25 billion with only 10% market penetration. When
13 including broader benefits and high penetration rates, AVs may save the U.S. economy roughly
14 $430 billion annually. While this does not include some associated costs and externalities, the
15 potential for dramatic change to the nature of transportation is very possible.
16
17 While potential benefits are substantial, significant implementation and mass-market penetration
18 barriers remain. Initial AV technology costs will likely be unaffordable for most households.
19 States are currently pursuing their own licensing and testing requirements, which may bring a
20 patchwork of regulations and requirements without federal guidance. An AV liability framework
21 is largely absent, creating uncertainty in the event of a crash. Security concerns should be
22 examined from a regulatory standpoint to protect the traveling public, and privacy issues must be
23 balanced against data uses. Car manufacturers have shown interest in AVs by investing millions
24 of dollars to make self-driving cars. The government should begin focusing research into how
25 AVs could impact transportation and land use patterns, and how to best alter our transportation
26 system to maximize their benefits while anticipating and mitigating negative impacts.
27
28 REFERENCES
29
30 American Automobile Association (2012) Your Driving Costs: How Much are you Really
31 Paying to Drive? Heathrow, FL.
32
33 Atiyeh, Clifford (2012) Predicting Traffic Patterns, One Honda at a Time. MSN Auto, June 25.
34
35 Berry, Irene (2010) The Effects of Driving Style and Vehicle Performance on the Real-World
36 Fuel Consumption of U.S. Light-Duty Vehicles. Massachusetts Institute of Technology.
37 Cambridge, MA.
38
39 Boesler, Matthew (2012) The 27 Best Selling Vehicles in America. Business Insider.
40
41 BMW of North America (2013) Build Your Own 2013 528i Sedan. Woodcliff Lake, NJ.
42
43 Brandon, John (2012) Privacy Concerns Raised over California “Robot Car” Legislation. Fox
44 News, September 14.
45

17
1 Bullis, Kevin (2011) How Vehicle Automation will Cut Fuel Consumption. MIT’s Technology
2 Review. October 24.
3
4 Bureau of Labor Statistics (2012) Occupational Outlook Handbook: Transportation and Moving
5 Occupations. Washington, D.C.
6
7 Campbell, Mark, Magnus Egerstedt, Jonathan How and Richard Murray (2010) Autonomous
8 Driving in Urban Environments: Approaches, Lessons and Challenges. Philosophical
9 Transactions of the Royal Society.
10
11 Cambridge Systematics (2011) Crashes vs. Congestion: What’s the Cost to Society? Prepared for
12 the American Automobile Association.
13
14 Center for Information and Society (2012) Automated Driving: Legislative and Regulatory
15 Action. Stanford, CA.
16
17 Cervero, Robert (2001) Induced Demand: An Urban and Metropolitan Perspective. Prepared for
18 Policy Forum: Working Together to Address Induced Demand. Berkeley, CA.
19
20 Dellenback, Steven (2013) Director, Intelligent Systems Department, Automation and Data
21 Systems Division, Southwest Research Institute. Communication by email, May 26.
22
23 Defense Advanced Research Projects Agency (2012) Grand Challenge ’05. Washington, D.C.
24
25 Dresner, Kurt, and Peter Stone (2008) A multiagent approach to autonomous intersection
26 management. Journal of Artificial Intelligence Research 31: 591-656.
27
28 Economist (2012) Watching your Driving, The Economist (June 23 issue): 80-81.
29
30 Economist Technology Quarterly (2012) Look, No Hands. September 1 issue: 17-19.
31
32 Farhadi, Ali, Ina Endres, Derek Hoiem, and David Forsyth (2009) Describing Objects by their
33 Attributes. 2009 Computer Society Conference on Computer Vision and Pattern Recognition.
34
35 Fagnant, Daniel and Kara Kockelman (2013). Environmental Implications for Autonomous
36 Shared Vehicles Using Agent-Based Model Simulation. Under review in Transportation Part C.
37
38 Federal Highway Administration (2005) Traffic Congestion and Reliability: Linking Solutions to
39 Problems. Washington, D.C.
40
41 Federal Highway Administration (2009) National Household Travel Survey. U.S. Department of
42 Transportation. Washington, D.C.
43
44 Federal Highway Administration (2013) Public Data for Highway Statistics. Office of Highway
45 Policy Information. Washington, D.C.
46

18
1 Grau, Alan (2012) President, Icon Labs. Telephone Interview, October 12.
2
3 Hayes, Brian (2011) Leave the Driving to it. American Scientist. 99: 362-366.
4
5 Hickey, Jason (2012) Vice President, Vínsula. Telephone interview, October 11.
6
7 J.D. Power and Associates (2012) 2012 U.S. Automotive Emerging Technology Study.
8
9 Kalra, Nidhi, James Anderson and Martin Wachs (2009) Liability and Regulation of
10 Autonomous Vehicle Technologies. California PATH Research Report UCB-ITS-PRR-2009-28.
11
12 Kockelman, Kara & Sukumar Kalmanje (2006) Road Pricing Simulations: Traffic, Land Use and
13 Welfare Impacts for Austin, Texas. Transportation Planning and Technology 29 (1): 1-23.
14
15 KPMG and CAR (2012) Self-Driving Cars: The Next Revolution. Ann Arbor, MI.
16
17 Kunze, Ralph, Richard Ramakers, Klaus Henning and Sabina Jeschke (2009) Organization of
18 Electronically Coupled Truck Platoons on German Motorways. Intelligent Robotics and
19 Applications: Second International Conference, Vol. 5928: 135-146.
20
21 LaHood, Ray (2011) Notice of Funding Availability for the Department of Transportation’s
22 National Infrastructure Investments Under the Full-Year Continuing Appropriations. U.S.
23 Department of Transportation. Federal Register 76 (156): 50310.
24
25 Litman, Todd (2012) Parking Management: Strategies, Evaluation and Planning. Victoria
26 Transport Policy Institute. Victoria, B.C.
27
28 National Highway Traffic Safety Administration (2008) National Motor Vehicle Crash
29 Causation Survey. U.S. Department of Transportation, Report DOT HS 811 059.
30
31 National Highway Traffic Safety Administration (2011) USDOT Connected Vehicle Research
32 Program: Vehicle-to-Vehicle Safety Application Research Plan. DOT HS 811 373.
33
34 National Highway Traffic Safety Administration (2012) Fatal Analysis Reporting System. U.S.
35 Department of Transportation, Washington D.C.
36
37 National Highway Traffic Safety Administration (2013) Traffic Safety Facts. U.S. Department of
38 Transportation, Washington D.C. DOT HS 811 753.
39
40 National Safety Council (2012) Estimating the Costs of Unintentional Injuries. Washington, D.C.
41
42 Newcomb, David (2012) Road-Train Test Keeps Cars in Line. Wired. May 29.
43
44 O’Brien, Chris (2012) Sergey Brin Hopes People will be Driving Google Robot Cars in “Several
45 Years”. Silicon Beat.
46

19
1 Puget Sound Regional Council (2006) 2006 Household Activity Survey. Seattle, WA.
2
3 Schrank, David and Tim Lomax (2011) 2011 Urban Mobility Report. Texas Transportation
4 Institute. College Station, TX.
5
6 Shladover, Steven, Dongyan Su and Xiao-Yun Lu (2012) Impacts of Cooperative Adaptive
7 Cruise Control on Freeway Traffic Flow. Proceedings of the 91st Annual Meeting of the
8 Transportation Research Board. Washington, D.C.
9
10 Smith, Bryant Walker (2012) Automated Vehicles are Probably Legal in the United States.
11 Center for Internet and Society. Stanford, CA.
12
13 State of Montana (2011) Montana Code Annotated 2011: 61-5-401. Driver License Compact.
14
15 Trottenberg, Polly (2011) Treatment of the Value of Preventing Fatalities and Injuries in
16 Preparing Economic Analysis – 2011 Revision. U.S. Department of Transportation, Washington
17 D.C.
18
19 Wood, Joanne (2002) Aging Driving and Vision. Clinical and Experimental Optometry; 85: 214-
20 220.
21
22 U.S. Census Bureau (2011) Age and Sex Composition: 2010. C2010BR-03.
23
24 U.S. House of Representatives and Senate (2012). MAP-21 Conference Report to Accompany
25 H.R. 4348. Report 112-557.

20