Automated & Autonomous Driving

Automated and
Autonomous Driving
Regulation under uncertainty
Corporate Partnership Board
Report
Corporate Partnership
Board
CPB

Automated and
Autonomous Driving
Corporate Partnership Board
Report

About the International Transport Forum
The International Transport Forum at the OECD is an intergovernmental organisation with 54 member
countries. It acts as a think tank for transport policy and organises the Annual Summit of transport
ministers. ITF is the only global body that covers all transport modes.
ITF works for transport policies that improve peoples’ lives. Our mission is to foster a deeper understanding
of the role of transport in economic growth, environmental sustainability and social inclusion and to raise
the public profile of transport policy.
ITF organises global dialogue for better transport. We act as a platform for discussion and pre-negotiation
of policy issues across all transport modes. We analyse trends, share knowledge and promote exchange
among transport decision makers and civil society.
Our member countries are: Albania, Armenia, Australia, Austria, Azerbaijan, Belarus, Belgium, Bosnia and
Herzegovina, Bulgaria, Canada, Chile, China (People’s Republic of), Croatia, Czech Republic, Denmark,
Estonia, Finland, France, Former Yugoslav Republic of Macedonia, Georgia, Germany, Greece, Hungary,
Iceland, India, Ireland, Italy, Japan, Korea, Latvia, Liechtenstein, Lithuania, Luxembourg, Malta, Mexico,
Republic of Moldova, Montenegro, Netherlands, New Zealand, Norway, Poland, Portugal, Romania, Russian
Federation, Serbia, Slovak Republic, Slovenia, Spain, Sweden, Switzerland, Turkey, Ukraine, United
Kingdom and United States.
Disclaimer
This report is published under the responsibility of the Secretary-General of the International Transport Forum. Funding for this
work has been provided by the ITF Corporate Partnership Board. This report has not been subject to the scrutiny of International
Transport Forum member countries. The opinions expressed and arguments employed herein do not necessarily reflect the official
views of member countries. This document and any map included herein are without prejudice to the status of or sovereignty over
any territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city or area.
www.internationaltransportforum.org

TABLE OF CONTENTS – 3
AUTOMATED AND AUTONOMOUS DRIVING: REGULATION UNDER UNCERTAINTY - © OECD/ITF 2015
Table of contents
Executive summary ......................................................................................5
Introduction ................................................................................................9
1. Automated and autonomous driving: Technologies,
contexts and capabilities.....................................................................11
Technologies and infrastructures.................................................................................. 11
Capabilities and contexts.............................................................................................. 13
2. Use cases, business cases and potential deployment pathways ................17
Products and services................................................................................................... 18
Deployment paths and technologies ............................................................................. 18
3. Regulatory environment .....................................................................25
Types of regulation....................................................................................................... 25
Regulatory approaches................................................................................................. 25
Regulatory considerations and policy choices ............................................................... 26
Key challenges ............................................................................................................. 28
Proposals for different automation strategies ............................................................... 29
Annex: Current regulatory frameworks for the testing of autonomous vehicles....30
Bibliography ..............................................................................................33
Figures
1. Technologies that allow vehicles to sense, plan and act in response to the dynamic driving environment ... 11
2. Levels of driving automation according to the Society of Automotive Engineers ....................................... 14
3. Transition from high to full driving automation across different contexts ................................................. 16
4. Urban mobility pathway from human to fully automated driving ............................................................ 21
5. Automated private vehicle pathway from human to fully automated driving ............................................ 22
6. Heavy-duty truck automation pathway from human to fully automated driving........................................ 23
7. Types of regulation for automated driving........................................................................................... 25

EXECUTIVE SUMMARY – 5
Executive summary
Many cars sold today are already capable of some level of automated operation, and prototype cars capable
of driving autonomously have been - and continue to be - tested on public roads in Europe, Japan and the
United States. These technologies have arrived rapidly on the market and their future deployment is
expected to accelerate. Autonomous driving promises many benefits: improved safety, reduced congestion
and lower stress for car occupants, among others. Authorities will have to adapt existing rules and create
new ones in order to ensure the full compatibility of these vehicles with the public’s expectations regarding
safety, legal responsibility and privacy. This report explores the strategic issues that will have to be
considered by authorities as more fully automated and ultimately autonomous vehicles arrive on our streets
and roads. It was drafted on the basis of expert input and discussions amongst project partners in addition
to a review of relevant published research and position papers.
What we found
Automated driving technologies are mostly mature and some autonomous driving is here already
Most of the core technologies required for fully autonomous driving are available today, many are mature
and some are already being deployed in commercially available vehicles.
Self-driving cars seem a near-term possibility but their range of capabilities is unclear
Many major car manufacturers and several technology firms have announced the commercial production of
highly automated vehicles starting in 2017. Many observers expect there to be a wide range of such models
on the market by 2030, and some of these may be self-driving. It is not clear at present, however, to what
extent these vehicles will be capable of self-driving in all circumstances.
Road safety is expected to improve with vehicle automation. But this effect remains untested at a large
scale and may not be immediate or linear
Most crashes involve human error. If greater autonomous operation reduces or eliminates these errors,
then benefits for road safety may be substantial. However, most driving involves no crashes. The real
safety test for autonomous cars will be how well they can replicate the crash-free performance of human
drivers. While results from early prototypes are promising, new types of crashes may emerge as
autonomous technologies become more common – for instance crashes resulting from the car handing
control back to the driver or from mixing autonomous and conventional vehicles.
There are many possible technological configurations for autonomous driving
The move towards autonomous driving may involve different technological configurations. Some rely on
greater connectivity between cars and between cars and infrastructure. These entail the development of
common communication protocols, encrypted security standards and investment in new types of
infrastructure or upgrading those which currently exist. Others rely more on vehicle-embarked sensor
platforms and require little infrastructure investment. Both models require precise digital representations of
their environment, including high definition maps.
There are two incremental paths towards full automation
The first path involves gradually improving the automation in conventional vehicles so that human drivers
can shift more of the dynamic driving task to these systems. The second path involves deploying vehicles
without a human driver in limited contexts and then gradually expanding the range and conditions of their
use. The first path is generally embraced by traditional car manufacturers and the second by new entrants.

6 – EXECUTIVE SUMMARY
Use and business cases are closely linked to automation pathways
Incremental advances in automation in conventional vehicles are unlikely to fundamentally change vehicle
market dynamics. It seems likely that individuals will buy and own such upgraded cars much as they do
today. Automated driving will be available for certain situations – for instance when driving on motorways,
parking a car, or handling stop-and-go traffic in case of congestion). Because the human driver must
resume active control when prompted to do so, such conditional automation raises particularly difficult
issues of human-machine interaction that have not been satisfactorily solved. Fully self-driving cars on the
other hand, will not face the same issue of human-machine coordination, although their use will likely be
confined to contexts where the vehicle can confidently handle the full range of driving complexity. Such
highly specific contexts include particular routes and low-speed operations. Self-driving cars have a much
higher potential for disruption. They may be deployed in fleet-wide systems that would fundamentally
reshape individual travel and have an impact on industries such as public transport and taxis.
Some regulatory frameworks are being developed for prototype testing, but not for future use cases
Several jurisdictions have passed or are considering rules that enable the testing, licensing and operation of
autonomous technologies and vehicles. Most address the safe operation of these vehicles on public roads,
although with little coordination amongst jurisdictions. We could not find evidence of anticipatory regulatory
action addressing the potential use cases that could result from large-scale deployment of highly
autonomous vehicles, such as the provision of quasi-public transport or taxi-like operations.
Policy insights
Automated driving comprises a diverse set of emerging concepts that must be understood individually and
as part of broader trends toward automation and connectivity
Vehicle automation is part of a much larger revolution in automation and connectivity. The recent hallmarks
of these revolutions – personal computers, mobile telephones, and the Internet – have converged and are
now blending with machines that sense and manipulate the physical environment. These machines include
not just automated motor vehicles but also drones, personal care robots, 3D printers, surveillance devices,
and many others. Vehicles will change with growing automation but so too will their role in society, and in
ways that are hard to foresee. Policies should account for this uncertainty and ensure sufficient resilience to
adapt to these changes or, at a minimum, not block changes that are desirable.
Uncertainty on market deployment strategies and pathways to automation complicates the regulatory task
Autonomous vehicle regulation should ensure safety and prevent, or at least mitigate, market failures. This
task is complicated by uncertainty on what it is that should be regulated and the risk that regulation may in
fact lock in one pathway to automation over a potentially better one. Though regulators may target
autonomous vehicles as a special case out of convenience, it may be preferable to adapt existing rules as
much as possible. While desirable, early regulatory action carries risks as well. Prematurely codifying
requirements can freeze unrealistic expectations – high or low – into the law in a way that causes the legal
framework to lag rather than to lead. Some regulatory flexibility seems desirable, for instance allowing
circumscribed uses such as low speed urban operation or motorway platooning before implementing a
blanket set of rules.
Incrementally shifting the driving task from humans to machines will require changes in insurance
Liability remains an important barrier for the manufacturers and designers of autonomous vehicles.
Expanding public insurance and facilitating greater private insurance could provide sufficient compensation
to those injured by an automated vehicle while relieving some of the pressure on the tort system to provide
such a remedy. Enhanced vehicle insurance requirements by manufacturers, especially if combined with
greater flexibility in the administration of this insurance, could also provide a third-party check on the

EXECUTIVE SUMMARY – 7
safety of automated systems. As automation increases, liability could gradually shift from drivers to
manufacturers and Original Equipment Manufacturers (OEMs). However, the allocation of liability among
these parties remains challenging and adjudication methods have yet to be developed.
The shift from human to machine may have an impact on what product information developers and
manufacturers of autonomous vehicles share, and with whom
Education of public actors, and of the public at large, is essential for developing effective regulations and
setting realistic expectations. Governments can facilitate public education by encouraging developers to
share specific data about their products and processes in order to benefit from more flexible regulation. In
some cases, it may be desirable to audit specific algorithms that directly impact public welfare, for instance
those that govern loss-loss decisions (such as the choice between two unavoidable crash scenarios) by
automated vehicles.
Regulators and developers should actively plan to minimise legacy risks
Vehicles with automated driving systems that are introduced in the next few years will not be perfect. Yet
they will still be present on roads for years after they have become outdated. An important goal for both
regulators and developers should be to limit the physical risk presented by legacy vehicles through a variety
of technical and contractual tools. These could include monitoring, over-the-air updates, and even virtual
recalls. Designing vehicles for future system upgrades like the addition of sensors may also help reduce
legacy risks.

INTRODUCTION – 9
Introduction
This report examines various emerging regulatory issues surrounding the deployment of automated and
autonomous vehicles. This work was based on the expert opinion of the authors and serves as a think piece
regarding the nature, timing and scope of regulatory action regarding automated and, ultimately, self-
driving vehicles.
The work for this report was carried out in the context of a project initiated and funded by the International
Transport Forum’s Corporate Partnership Board (CPB). CPB projects are designed to enrich policy discussion
with a business perspective. They are launched in areas where CPB member companies identify an
emerging issue in transport policy or an innovation challenge to the transport system. Led by the ITF, work
is carried out in a collaborative fashion in working groups consisting of CPB member companies, external
experts and ITF researchers.
The principal authors of this report were Dr Bryant Walker Smith, Assistant Professor at the University of
South Carolina and affiliate scholar at the Center for Internet and Society at Stanford Law School, and
Joakim Svensson, Director at Volvo Group. Substantial inputs were provided by Paulo Humanes of PTV
AG, Gilbert Konzett and Alexander Paier of Kapsch TrafficCom, Takuma Mizuno of Nissan and Antigoni
Lykotrafiti of ITF, who undertook valuable research in support of the work. The report was edited by
Philippe Crist.
The project was coordinated by Philippe Crist and Sharon Masterson of the International Transport Forum.

AUTOMATED AND AUTONOMOUS DRIVING – 11
1. Automated and autonomous driving: Technologies,
contexts and capabilities
Automated driving encompasses a wide range of technologies and infrastructures, capabilities and contexts,
use cases and business cases, and products and services. There is no single timeline for these
developments: Some are here today, some may be distant, and some will depend on specific technical
innovations or particular policy choices. This section surveys these potential developments in order to
inform the policy choices that could enable or constrain them.
Importantly, vehicle automation is part of much larger revolutions in automation and connectivity (Smith,
2014b). The recent hallmarks of these revolutions – personal computers, mobile telephones, and the
Internet – have converged with each other and are now blending with machines that sense and manipulate
the physical environment. These machines include not just automated motor vehicles but also drones,
personal care robots, 3D printers, surveillance devices, and many others. While addressing only vehicle
automation, this report strives to anticipate these broader changes in both technology and society.
Technologies and infrastructures
An automated vehicle, like a human, must collect information, make a decision based on that information,
and execute that decision. Information comes from vehicle equipment, physical infrastructure, physical-
digital infrastructure, and digital infrastructure, any of which may be public or private. Many of these
technologies exist today and are capable of guiding vehicles and in some cases drive vehicles with minimal
or no driver input in test situations and across diverse driving environments.
Figure 1. Technologies that allow vehicles to sense, plan and act in response to
the dynamic driving environment
Source: Adapted from (Parrish, 2015).
Adaptive
cruise control
Emergency braking
Cross traffic
warning
Surround view
Surround view
Parkassist
Parkassist
Blind
spot
detection
Rear
collision
warning
Long-range radar LIDAR (laser scan) Camera Short/medium-range radar Ultrasound
HighRes Map
Processors
Secure coms.
Algorithms
Sense
Plan
Act

12 – AUTOMATED AND AUTONOMOUS DRIVING
Many companies have carried out trials or are engaged in continuous on-road testing of highly automated
vehicle prototypes whose capacities are evolving rapidly due to improved sensor-processing technologies,
adaptive algorithms, high-definition mapping and in some cases, the deployment of vehicle-to-vehicle and
infrastructure-to-vehicle communication technologies. Automated vehicles have crossed the United States
almost solely in self-driving mode and have undertaken long-distance motorway and arterial trips in Europe
and Japan.
From a technical point of view, current technology for highly automated driving in controlled environments
is quite mature. These vehicles uses state-of-the-art sensors (radar, lidar, GPS and camera vision systems)
combined with high accuracy maps allowing on-board systems to identify appropriate navigation paths, as
well as obstacles and relevant signage. These prototypes operate with a driver that must stand ready to
take control of the vehicle though reports from trials indicate that this option is rarely acted upon. As of
2015, there is yet no consensus on the commercial maturity of highly automated and ultimately fully
automated driving. Some manufacturers have announced the arrival of highly automated and possibly fully
automated vehicles by as early as 2016 while others have advanced much later dates (up to 2030). Clearly
there will be a first-mover advantage for pioneers in the field of highly automated driving but there are also
risks linked to the safety performance of these vehicles and the possibility of regulatory action that inhibits
technology development and deployment.
A number of issues will have to be addressed in order to support the deployment of high-automation
scenarios. These include:
 Vehicle-to-X connectivity (V2X): Connectivity is an important element of the automated
vehicles especially secure V2X communication requiring low latency. V2X technologies encompass
the use of wireless technologies to achieve real-time two-way communication among vehicles
(V2V) and between vehicles and infrastructure (V2I). The convergence of sensor-based solutions
(current advanced driver assistance - ADAS) and V2X connectivity will promote automated driving.
 Decision and control algorithms: These include decision, planning and control algorithms for a
cooperative, safe, human compatible traffic automation.
 Digital infrastructure: Digital infrastructure (for road automation) includes static and dynamic
digital representations of the physical world with which the automated vehicle will interact to
operate. Issues to address include: sourcing, processing, quality control and information
transmission.
 Human factors: Human factors in automation relate to understanding the interaction(s) of
humans with all aspects of an automated road transport system (ART), both from within a vehicle,
when taking the role of a driver/operator and also as a road user, when interacting with automated
vehicles. Knowledge and theories from social-psychological and behavioural sciences are useful to
understand how humans interact with such systems.
 Evaluating road automation: Automation of road vehicles has the potential to impact on
lifestyles and society. Economic impacts too will be important and it will be necessary to gauge
these impacts in a common cost-benefit framework with other transport investments when
assessing public expenditure on supporting infrastructure or services.
 Roadworthiness testing: Roadworthiness testing, understood as the necessary tests to evaluate
if a vehicle is legally allowed to drive on public roads is of capital importance for the deployment of
new automated driving functionalities.

Capabilities and contexts
The wide variety of relevant technologies has inspired complex – and indeed competing – taxonomies of
vehicle automation. Of these, International Society of Automotive Engineers’ (SAE) Levels of Driving
Automation capture the emerging descriptive consensus most systematically and are accordingly presented
in detail here. Figure 2 (SAE, 2014) summarises these levels. The expert committee responsible for this
taxonomy emphasises:
“These levels are descriptive rather than normative and technical rather than legal. They imply no particular
order of market introduction. Elements indicate minimum rather than maximum system capabilities for
each level. A particular vehicle may have multiple driving automation features such that it could operate at
different levels depending upon the feature(s) that are engaged.” (SAE, 2014)
As with other approaches, these levels primarily identify how the “dynamic driving task” is divided between
human and machine: It is performed entirely by a human driver at Level 0 (no automation) and entirely by
an automated driving system at Level 5 (full automation). In the “mushy middle,” (Smith, 2014a) this task
is shared simultaneously or sequentially, raising difficult questions of human-machine interaction.
Current deployment and development necessarily focus on this middle. This is because full automation, “the
full-time performance by an automated driving system of all aspects of the dynamic driving task under all
roadway and environmental conditions that can be managed by a human driver,” (SAE, 2014) remains
elusive. Human drivers confront, and usually manage, an incredible variety of contexts—geographic areas,
roadway types, traffic conditions, weather conditions, and events/incidents—for which automated vehicles
have yet to be designed and demonstrated.
Efforts toward full automation tend to follow one of two incremental paths. The first involves gradually
improving the automated driving systems available in conventional vehicles so that human drivers can shift
more of the dynamic driving task to these systems. The second involves deploying vehicles without a
human driver and gradually expanding this operation to more contexts. These two approaches can be
simplistically described as “something everywhere” and “everything somewhere.”
The “something everywhere” strategy is generally embraced by traditional car manufacturers and is well
captured by the levels of automation. Many of today’s production vehicles are capable of driver assistance
(Level 1), typically through the use of adaptive cruise control to adjust speed based on following distance.
A small number of vehicles also incorporate an active lane-keeping assist feature in a way that makes them
capable of partial automation (Level 2). Notwithstanding the potential for and reality of driver distraction,
both of these levels assume that the human driver continues to actively monitor the driving environment.

Figure 2. Levels of driving automation according to the Society of Automotive Engineers
Source: Adapted from SAE Standard J3016 (SAE, 2014).
The introduction of conventional cars and trucks capable of operating without this active monitoring will
represent a significant technical and conceptual leap. This threshold between partial automation (Level 2)
and conditional automation (Level 3) corresponds to the line that several US states have drawn between
non-automated and automated vehicles. Because of its assumption that the human driver will resume
actively driving shortly after being prompted to do so, conditional automation raises particularly difficult
issues of human-machine interaction that have not been satisfactorily solved. A more normative taxonomy
developed by the U.S. National Highway Traffic Safety Administration arguably passes over conditional
automation altogether on its way to high automation (NHTSA, 2013).
SAE
Level
Name
Steering,
acceleration,
deceleration
Monitoring
driving
environment
Fallback
performance
of dynamic
driving task
System
capability
(driving
modes)
Humanmonitorsenvironment
0
No automation
the full-time performance by the human
driver of all aspects of the dynamic driving
task, even when enhanced by warning or
intervention systems
1
Driver assistance
the driving mode-specific execution by a
driver assistance system of either steering or
acceleration/deceleration using information
about the driving environment and with the
expectation that the human driver perform
all remaining aspects of the dynamic driving
task.
Some driving
modes
2
Partial automation
the driving mode-specific execution by one
or more driver assistance systems of both
steering and acceleration/deceleration using
information about the driving environment
and with the expectation that the human
driver perform all remaining aspects of the
dynamic driving task
Some driving
modes
Carmonitorsenvironment
3
Conditional automation
the driving mode-specific performance by an
automated driving system of all aspects of
the dynamic driving task with the
expectation that the human driver will
respond appropriately to a request to
intervene
Some driving
modes
4
High automation
the driving mode-specific performance by an
automated driving system of all aspects of
the dynamic driving task, even if a human
driver does not respond appropriately to a
request to intervene
Some driving
modes
5
Full automation
the full-time performance by an automated
driving system of all aspects of the dynamic
driving task under all roadway and
environmental conditions that can be
managed by a human driver
All driving
modes

Levels of automation beyond conditional automation (Level 3) may operate on the basis of inputs solely
from vehicle-embarked sensors (self-sensing) or via a combination of self-sensor input and inputs from
sensors embarked on other vehicles and infrastructure that are communicated to the vehicle to the vehicle
in near real-time. The connected vehicle and connected infrastructure approach requires available data
transmission frequencies, low-latency, trusted, secure and fail-safe data transmission protocols and
harmonised data syntax that ensures safe interoperability. Work is underway to address all three of these
requirements for connected vehicle operation and early efforts, like the development of a co-operative ITS
corridor testbed reaching from the Netherlands through Germany to Austria – the ECoAT project (see
box 1) – will lay the groundwork for future deployment of connected vehicles. The United States is similarly
developing testbeds and trialling connected service frameworks, but it is not clear at present to what extent
future autonomous vehicles will be connected or truly autonomous (relying minimally on external data
inputs).
Box 1. ECoAT Cooperative ITS corridor (Netherland, Germany, Austria)
The ECoAT project provides a basis for standardised, international, future-oriented cooperative ITS services.
These will be useful as vehicles become more automated and will likely spur at least partial automation in areas
where these services are fully deployed. The project centres on the deployment of cooperative ITS services on a
testbed corridor linking the Netherlands, Germany and Austria. The project sketches out a joint road map for the
introduction of initial cooperative ITS services, promotes the development of connected ITS services by agreeing
common functional descriptions of an initial set of cooperative ITS services as well as their technical specifications
and frames the implementation of these services along the testbed corridor.
In its first phase, ECoAT will develop and deploy two cooperative ITS services: one based on standardised and
machine-readable and interpretable warning data regarding road works and another service seeking to improve
traffic management by exploiting vehicular and infrastructure data. These and further services yet to be
developed by the ECoAT living lab will contribute to a comprehensive systems specification for co-operative ITS
services developed by national authorities and industrial partners.
High automation (Level 4) is nonetheless challenging because it describes an automated driving system
that, once engaged, can always revert to a “minimal risk condition” should a human driver not resume
actively driving. Reverting to this minimal risk condition may be easier in some contexts (e.g. low-speed
parking facilities) than in others (e.g. urban expressways) For this reason, a highly automated driving
system is capable of operating in some, but not necessarily all, contexts or “driving modes”.
High automation (Level 4) is where the “everything somewhere” strategy begins. The custom vehicles that
currently operate without any real-time input from human drivers are limited to highly specific contexts,
including particular routes and low speeds. Examples include the CityMobil2 project supported by the
European Union, the Navia shuttle marketed by Induct, and the buggies announced by Google. Such fully
automated vehicles have also been deployed in restricted contexts for freight or industrial tasks, including
tyre-based container repositioning vehicles in ports or fully automated ore trucks in some open-air mines.
A key challenge is introducing these vehicles to more geographic areas, roadway types, traffic conditions,
weather conditions, and events/incidents. One developer, for example, might initiate a pilot project in which
his vehicles operate in good weather at neighbourhood speeds along a carefully mapped, maintained, and
monitored corridor within its corporate campus. It might then expand the pilot to select streets within the
local community and later to a handful of other communities. As the developer improved its technologies
and increased public confidence in them, it might deploy vehicles at higher speeds and on more road types.
Such a system of automated vehicles might eventually function in many traffic and weather conditions on
many roads in many communities. Nonetheless, these vehicles would not reach full automation (Level 5)

unless they handled “all roadway and environmental conditions that can be managed by a human driver.”
Figure 3 illustrates this long path from high automation to full automation across different driving modes.
One outcome of this context-conditional performance spectrum is that regulatory authorities may issue
context-dependent operating licenses matching vehicle classes and specific contexts (e.g. non-motorway
operation with maximum speeds of no more than 40 km/hr). At present though, it is not yet clear whether
authorities will seek to follow this vehicle-context licensing path and if so, whether this would only be a
short-term phenomenon or a more permanent aspect of automated vehicle regulation. What seems clear is
that this approach would be easier to deploy on the basis of specified performance characteristics (safety,
speed, road class), rather than on the specification of certain technologies.
Figure 3. Transition from high to full driving automation across different contexts

USE CASES, BUSINESS CASES AND POTENTIAL DEPLOYMENT PATHWAYS – 17
2. Use cases, business cases and potential deployment
pathways
The two approaches outlined in the previous section correspond to different use cases and business cases.
Whereas freeways may be the most promising early application for increasingly automated conventional
cars and trucks, urban areas are well-suited for specialised passenger and delivery shuttles.
The “something everywhere” strategy for conventional cars and trucks points to increasingly sophisticated
advanced driver assistance systems (ADASs) that are marketed in terms of safety, fuel efficiency, driver
comfort, driver convenience, and ultimately driver productivity (assuming these benefits are realised).
Automated emergency intervention systems (AEISs), which momentarily perform all or part of the dynamic
driving task to avoid an imminent crash but which are excluded from SAE’s taxonomy, are also likely to be
offered in combination with advanced driver assistance systems.
Motorway automation may be an early use case for conditional or high automation in conventional vehicles.
Although speeds are high, motorways tend to be more uniformly designed and better maintained. Vehicle
flows are more organised, and bicyclists and pedestrians are generally absent. There are two relevant notes
on motorway automation: dedicated facilities and vehicle platoons.
Dedicated facilities are occasionally proposed for automated freeway vehicles, but retrofitting existing
facilities is likely to be prohibitively expensive and may ultimately prove unnecessary. Separation may be
more viable on newly constructed roadways in rapidly urbanising countries, on existing managed lanes
(such as those for high-occupancy vehicles) between major employment and residential areas, and on
specialised facilities serving exceptionally large numbers of trucks.
Vehicle platoons are a particularly promising application for freeways. As typically envisioned, a platoon
consists of two to six cars or trucks that are closely spaced and tightly coordinated through both vehicle-to-
vehicle communication and some degree of automation. A driver may sit in each vehicle, in only the lead
vehicle, or eventually in none of the vehicles. Benefits may include significant fuel savings and, for fleet
operators, potentially lower labour costs.
Vehicle automation systems and especially automated emergency intervention systems may also have early
applications beyond freeways. Automation may be appropriate for low-speed travel during peak periods.
Particular parking facilities may support automated valet functions. And conventional cars assigned to car
sharing programmes might eventually reposition themselves by traveling at low speeds on particular roads
during non-peak periods.
Many urban and suburban applications, however, might be realised earlier through an “everything
somewhere” strategy of nonconventional vehicles. Passenger shuttles and taxis might operate at low
speeds in central business districts, corporate campuses, university campuses, military bases, retirement
communities, resorts, shopping centres, airports, and other semi closed environments as well as for first-
and last-mile transit applications.
Delivery shuttles might likewise travel at low speeds along particular routes and at particular times.
Depending on their size and purpose, these robotic delivery systems might conceivably use pathways rather
than or in addition to roadways. Indeed, the potential proliferation of service robots might bring a new kind
of nonhuman user to the urban environment.
Some of these urban applications may benefit from specialised infrastructure. Physical infrastructure might
include vehicle-to-vehicle and vehicle-to-infrastructure communications equipment, ground-based units for
global navigation systems, dedicated facilities comparable to bus and bicycle lanes, on-street parking
restrictions, and specific roadway or pavement modifications. Digital infrastructure might include the

18 – USE CASES, BUSINESS CASES AND POTENTIAL DEPLOYMENT PATHWAYS
maintenance of highly detailed roadway maps and pertinent traffic operations data. This specialised
infrastructure, if actually required, could be limited to a manageable set of corridors actually used by a
particular urban mobility system.
Whereas wealthy consumers and fleet operators are likely to be early adopters of “something everywhere”
vehicles, an “everything somewhere” approach might reach a more diverse group of users. Especially if its
fuel and labour costs are lower and its usage is higher, an extensive urban mobility system might compare
favourably with private vehicle ownership, conventional taxis, and conventional public transit. Residents
who cannot afford to buy and maintain a private car or who are unable to drive may be some of the earliest
adopters of these shared systems.
Products and services
Vehicle automation could give rise to particular consumer-oriented products and services. The “something
everywhere” strategy for conventional cars and trucks is likely to rely primarily on a traditional model of
selling and especially leasing vehicles to individual consumers or fleet operators. However, manufacturers
are likely to be more closely connected to the owners and users of their vehicles through a variety of
contractual and technical tools (Smith, 2014b). These may include terms of use, end-user license
agreements, and subscription agreements on the contractual side and advanced telematics, driver
monitoring, and over-the-air updates on the technical side. Manufacturers might also pursue new revenue
streams via automation subscription services, consumer-facing advertising, or the marketing of user data.
In addition, companies other than car manufacturers may seek to add, enhance, or customise vehicle
automation systems through aftermarket conversions and modifications. Production vehicles already
provide a platform for many automated vehicle research efforts, at least one start-up company has
announced its intention to add a partial automation feature to certain production vehicles, and legislatures
in several US states have expressly limited a car manufacturer’s civil liability for injuries caused by a third-
party’s addition of automated driving technology to one of its production vehicles.
An “everything somewhere” strategy could more fully embrace a variety of service models. Passenger
shuttles, automated taxis, delivery services, and other urban concepts are likely to involve some kind of
central ownership, management, maintenance, and dispatch. These services may be public, private, or
hybridised, and they may complement or compete with conventional public transit.
Deployment paths and technologies
As noted above, deployment pathways will follow either an incremental evolution of traditional vehicles
leading to higher and higher levels of automation or a radical technology-shift approach that would lead to
the near-term deployment of highly automated urban mobility vehicles. These pathways are illustrated in
Figures 4, 5 and 6. The remainder of this section describes the technical aspects relating to the deployment
of these pathways without pre-judging their likelihood or impact.
All pathways build on the range of automated driving technologies and systems already commercially
deployed today. These include:
Current and future vehicle systems on Level 0 (no automation)
 Systems beyond human capability to act: There are several systems on the market today that
intervene beyond the human capability to act. These systems, like ABS (Anti-Lock System), ESC
(Electronic Stability Control) and emergency braking are active safety systems that allow higher
levels of automation and will facilitate deployment. Future version of these systems will include
emergency evasion and emergency stopping.

 Lane Change Assist (LCA): The system monitors the areas to the left and right of the car and up
to 50 metres behind it and warns the driver of a potentially hazardous situation by means of
flashing warning lights in the exterior mirrors.
 Park Distance Control (PDC): The Park Distance Control system assists the driver to manoeuvre
into tight spaces and reduces stress by communicating distance from obstacles by means of
acoustic or, depending on vehicle, optical signals.
 Lane Departure Warning (LDW): Lane Departure Warning helps to prevent accidents caused by
unintentional wandering out of traffic lanes. It represents a major safety gain on motorways and
major trunk roads. If there is an indication that the vehicle is about to leave the lane
unintentionally, the driver is alerted visually and in some cases by a signal on the steering wheel.
 Front Collision Warning (FCW): The Front Collision Warning monitoring system uses a radar
sensor to detect situations where the distance to the vehicle in front is critical and helps to reduce
the vehicle’s stopping distance. In dangerous situations the system alerts the driver by means of
visual and acoustic signals and/or with a warning jolt of the brakes. Front Collision Warning
operates independently of the adaptive cruise control or automatic distance control.
Current and future vehicle systems on Level 1 (driver assistance)
 Adaptive Cruise Control (ACC): The cruise control system with “automatic distance control ACC”
uses a distance sensor to measure the distance and speed relative to vehicles driving ahead. The
driver sets the speed and the required time gap with buttons on the multifunction steering wheel or
with the steering column stalk (depending on model). The target and actual distance from following
traffic can be shown as a comparison in the multifunction display.
 ACC including stop-and-go function: Adaptive cruise control with stop and go function includes
automatic distance control (control range 0–250 km/h) and, within the limits of the system, detects
a preceding vehicle. It maintains a safe distance by automatically applying the brakes and
accelerating. In slow-moving traffic and congestion it governs braking and acceleration.
 Lane Keeping Assist (LKA): Lane Keeping Assist automatically becomes active from a specific
speed (normally from around 60 km/h) and upwards. The system detects the lane markings and
works out the position of the vehicle. If the car starts to drift off lane, the LKA takes corrective
action. If the maximum action it can take is not enough to stay in lane, or the speed falls below 60
km/h, the LKA function warns the driver, for instance with a vibration of the steering wheel. It is
then for the driver to take correcting action.
 Park Assist (PA): The Park Assist function automatically steers the car into parallel and bay
parking spaces, and also out of parallel parking spaces. The system assists the driver by
automatically carrying out the optimum steering movements in order to reverse-park on the ideal
line. The measurement of the parking space, the allocation of the starting position and the steering
movements are automatically undertaken by Park Assist – all the driver has to do is operate the
accelerator and the brake. This means that the driver retains control of the car at all times.
Urban mobility pathway
This pathway encompasses the types of initially low-speed, fully automated but limited operation vehicles
that could be deployed in urban areas. Current high automation systems have been deployed in limited
areas or on dedicated infrastructure. This will be the base for going to higher and higher vehicles speeds
and perhaps less specific requirements on the infrastructure. Possible use cases include:

 Cyber cars, cyber vans, cyber minibuses. These are small-to-medium-sized automated vehicles
for individual or collective transport of people or goods with the following characteristics: a) They
are fully automated on demand transport systems that under normal operating conditions do not
require human interaction; b) they can be fully autonomous or make use of information from a
traffic control centre, information from infrastructure or information from other road users; c) they
are small vehicles, either for individual transport (1-4 people) or for transport of small groups (up
to 20 people); d) they can either use a separated infrastructure or a shared space.
 High-Tech Buses. These are buses on rubber wheels, operating more like trams than like
traditional buses, with the following characteristics: a) They are vehicles for mass transport (more
than 20 people); b) they use an infrastructure, which can be either exclusive for the buses or
shared with other road users; c) they can use various types of automated systems, either for
guidance or for driver assistance; d) they always have a driver, who can take over control of the
vehicle at any time, allowing the vehicles to use the public road.
 Personal Rapid Transit (PRT). This is a transport system featuring small fully automatic vehicles
for the transport of people, with the following characteristics: a) PRT systems operate on its own
exclusive infrastructure, so there is no interaction with other traffic; b) they are fully automated
systems that under normal operating conditions do not require human interaction; c) they are
small with a capacity usually limited to 4 to 6 persons per vehicle; d) PRT offer on-demand service,
where people are transported directly from origin station to destination station without stopping at
intermediate stations, without changing vehicles and ideally without waiting time.
 Advanced City Cars (ACC). New city vehicles integrating zero or ultra-low pollution mode and
driver assistance such as ISA (Intelligent Speed Adaptation), parking assistance, collision
avoidance, stop-and-go function, etc. These vehicles should also incorporate access control coupled
with advanced communications in order to integrate them easily into car-sharing services.
 Dual-mode vehicles. Developed from traditional cars but able to support both fully automatic and
manual driving. The first applications of automatic driving will be for relocation of shared cars using
platooning techniques, but dual-mode vehicles could become full cyber cars in specific areas or
infrastructures. They represent the migration path from traditional cars to automatic driving.
Figure 4 below illustrates a potential pathway for automated urban mobility systems. This pathway
encompasses both cyber cars (including cyber vans and cyber minibuses) and automated buses and
personal rapid transit systems.
With first-generation cyber cars (Level 4), the last mile taxi or delivery vehicle is fully automated in its area
of operations, taking a limited number of passengers with a maximum speed of 40 km/h. It operates in a
specific area with dedicated infrastructure. For second-generation cyber cars (Level 4), the last mile taxi or
delivery vehicle is fully automated in its area of operations taking a limited number of passengers. It
operates in a specific area with adapted infrastructure. Automated taxis (Level 5) offer fully automated
driving that can in principle take its passenger to all places. It should be noted that no realistic time
estimation for the availability of this system exits.
For automated buses or Personal Rapid Transit, the pathway begins at Level 4 with first-generation vehicles
driving in segregated lanes and using dedicated infrastructure, with a maximum speed of 40km/h. Second-
generation automated buses and PRT (Level 4) use dedicated bus lanes and supporting infrastructure at
normal city vehicle speeds. Additional functionality such as adaptive urban traffic control systems that
control the traffic lights and gives speed advice and priority can be introduced when these systems reach
the market. At the next stage, still at Level 4, an automated bus drives in mixed traffic in the defined area
of operation.

Figure 4. Urban mobility pathway from human to fully automated driving
Source: Based on CityMobil2 project (CityMobil2, 2015).
Automated private vehicle pathway
Figure 5 illustrates a potential pathway for the automation of private individually owned vehicles. This
pathway leads from existing commercially-deployed systems to a fully self-driving car in incremental steps.
 Automated Parking Assistance: Automated parking assistance is available on the market today.
 Park Assist (Level 2): Partial automated parking into and out of a parking space in a public or
private parking area or garage. The process is initiated remotely, e.g. via smartphone or adapted
remote key. The vehicle carries out the manoeuvre by itself. The driver can be located outside of
the vehicle, but has to monitor the system and can stop the parking manoeuvre if required.
 Parking Garage Pilot (Level 4): Highly automated parking including manoeuvring to and from
parking place (driverless valet parking). In parking garages, the driver does not have to monitor
the operation and may leave once the system is active. The process is initiated remotely, for
instance via a smartphone or an adapted remote key.
 Traffic Jam Assist (Level 2): The function controls the forward/backward and sideways movement
of the vehicle in order to follow traffic flow in low speeds below 30 km/h. The system can be seen
as an extension of the ACC with stop-and-go functionality.
 Traffic Jam Chauffeur (Level 3): Conditional automated driving in congested conditions up to 60
km/h on motorways and motorway-like roads. The system controls the forward/backward and
lateral movements of the vehicle up to the threshold speed. The driver must deliberately activate
Level 0
No driving task
automation
(passive driver assistance
and advanced driver
assistance system beyond
human capability to act –
e.g. ABS braking)
Level 1
Active driver
assistance
Level 2
Partial automation
Level 3
Conditional
automation
Level 4
High automation
Level 5
Full automation
Established technology
2015
2030
Lane change assist (LCA)
Park distance control (PDC)
Lane departure
warning (LDW)
Front collision
warning (FCW)
Emergency braking
Anti-lock braking (ABS)
Electronic stability control (ESC)
Driver steering recommendation (DSR)
Adaptive cruise control (ACC) +
ACC + Stop & Go (SG)
Lane keeping assist (LKA)
Basic park assist (PA)
Lane departure warning (LDW)
Cybercars/
delivery vehs.
Last mile use, and
Automated
bus/PRT in
dedicated lane
Cybercars
Self-driving “taxi”
or delivery van in
mixed traffic
Automated
bus/PRT in mixed
traffic
Cybercars/
delivery vehs.
Last mile use, low
speed context and
Automated
bus/PRT:
segregated lane
Urban Mobility Pathway

the system, but does not have to monitor the system constantly. The driver can override or switch
the system off at all times. There is no take over request to the driver from the system.
 Highway Chauffeur (Level 3): Conditional automated driving up to 130 km/h on motorways or
motorway-like roads. The Highway Chauffeur operates from entrance to exit, on all lanes, including
overtaking movements. The driver must deliberately activate the system, but does not have to
monitor it constantly. The driver can override or switch off the system at all times. The system can
request the driver to take over within a specific time, if automation reaches the system limits.
 Highway Pilot (Level 4): Automated driving up to 130 km/h on motorways or motorway-like
roads from entrance to exit, on all lanes, including overtaking movements. The driver must
deliberately activate the system, but does not have to monitor it constantly. The driver can
override or switch off the system at all times. There are no requests from the system to the driver
to take over when the system is in its normal operation area on the motorway. Depending on the
deployment of vehicle-to-vehicle communication and cooperative systems, ad-hoc convoys of
vehicles (platoons) could also be created.
 Fully automated private vehicle (Level 5): The fully automated vehicle should be able to handle
all driving from point A to B, without any input from the passenger. The driver can at all times
override or switch off the system. No consensus exists as to when such systems will become
commercially available.
Figure 5. Automated private vehicle pathway from human to fully automated driving
Level 0
No driving task
automation
and advanced driver
e.g. ABS braking)
Level 1
Active driver
assistance
Level 2
Partial automation
Level 3
Conditional
automation
Level 4
High automation
Level 5
Full automation
2015
2030
Lane departure
warning (LDW)
Front collision
warning (FCW)
Emergency braking
Partial automated
parking and
Traffic jam assistance
Traffic jam chauffeur
Highway chauffeur
Parking garage pilot
Highway autopilot
including highway convoy
Self-driving vehicle
Automated Private Vehicle Pathway

Truck automation pathway
Figure 6 below outlines an automation pathway for heavy-duty trucks. One of the key automation strategies
here will be the deployment of truck platoons or “road trains”. These could include not only other trucks but
could also serve to tether individual vehicles in mixed truck and car platoons.
Figure 6. Heavy-duty truck automation pathway from human to fully automated driving
 CACC Platooning: Partially automated truck platooning in which trucks are coupled by cooperative
ACC (CACC). Engine and brake control keeps a short but safe distance to the lead vehicle. Drivers
remain responsible for all other driving functions.
 Truck platooning: This function enables platooning in a specific lane. The vehicle should be able
to keep its position in the platoon with a fixed distance or fixed time difference from the front
vehicle. The behaviour of the first vehicle, such as braking and steering, should be transmitted by
vehicle-to-vehicle communication. The function should also smoothly handle vehicles leaving the
platoon. Dedicated “perpetual” truck platoons that allow vehicles to join and leave at specified
stations could also prove an interesting future application of this technology for long-distance
motorway trips. Up-scaling and deployment can be reached as follows:
‐ Start with trucks, as there is a strong financial incentive due to 10% to 15% fuel savings.
‐ Start with small platoons of only two trucks and co-operation with fleet-owners in high density
trucking areas.
‐ For legal reasons, start with a system where the following truck still has a driver in it.
‐ Set up an (open) fleet management system for trip matching between equipped trucks of
different fleet owners.
Level 0
No driving task
automation
and advanced driver
e.g. ABS braking)
Level 1
Active driver
assistance
Level 2
Partial automation
Level 3
Conditional
automation
Level 4
High automation
Level 5
Full automation
2015
2030
Lane departure
warning (LDW)
Front collision
warning (FCW)
Emergency braking
Traffic jam assistance
Truck platooning, and
Traffic jam chauffeur
Highway chauffeur
Truck terminal parking
Highway autopilot
with ad-hoc platooning
Self-driving trucks
Truck Automation Pathway
Co-operative ACC,
platooning

 Traffic Jam Assist (Level 2): Identical function as for passenger vehicles (see previous section).
 Traffic Jam Chauffeur (Level 3): Identical function as for passenger vehicles (see previous
section).
 Highway Chauffeur (Level 3): Identical function as for passenger vehicles (see previous section).
 Highway Pilot with ad-hoc platooning (Level 4): In addition to the functions described for
passenger vehicles above, ad-hoc convoys could also be created. This requires availability of
vehicle-to-vehicle communication and depends on the deployment of cooperative systems.
 Fully automated truck (Level 5): The fully automated truck should be able to handle all driving
from point A to B. The driver can at all times override or switch off the system. No consensus exists
as to when such systems will become commercially available.

REGULATORY ENVIRONMENT – 25
3. Regulatory environment
Types of regulation
Encouraging desirable conduct and deterring undesirable conduct are among the most important goals of
regulation. In a narrow sense, regulation is simply the enactment of a binding rule by a public authority.
This narrow definition, however, denies other key regulatory tools that warrant attention. Figure 7 arranges
several of these tools according to whether the actor wielding them is public or private and whether the
action taken is forward looking (ex ante) or backward looking (ex post).
Any of these tools could impact whether automated driving systems are deployed and, if they are, what
kinds, when, where, how, and by whom. For example, imposing an insurance requirement on developers of
automated vehicles, as several US states have now done, could advantage larger companies that self-
insure as well as private insurers to whom smaller developers may turn.
Figure 7. Types of regulation for automated driving
Source: Smith B. W., 2013.
Regulatory approaches
Today’s motorists and motor vehicles already pass through a number of regulatory gates. For transport
products, particularly non-commercial vehicles, these can include first sale (which typically requires
manufacturers to either self-certify or to obtain type approval), vehicle registration and subsequent renewal
by the owner, driver licensing and subsequent renewal for the operator, provision of vehicle insurance,
traffic and vehicle enforcement, investigations into vehicle defects, recall of vehicles or components that are
not reasonably safe, and litigation over traffic injuries and fatalities.

26 – REGULATORY ENVIRONMENT
For transport services, including public transit, taxicab operations, and commercial trucking, these gates
can include all of those listed above as well as construction of facilities, procurement of rolling stock,
awards of concessions, professional drivers licensing and funding of projects and programmes.
These gates enable governments to monitor, influence, or impede particular vehicle automation concepts.
They may also obligate public actors to address automation sooner than they are ready. Consider the
following examples:
 A police officer observes a driver operating a vehicle without touching the steering wheel.
 An agency charged with registering only those vehicles that are “reasonably safe” receives an
application for a vehicle with automation capabilities that have been added by its owner.
 A driver licensing agency receives an application from a disabled person who cannot legally drive
under existing rules but who asserts the right to operate an automated vehicle under an
accommodations law.
 A transport agency that has completed an environmental review for a major infrastructure project
faces a legal challenge by an opponent of the project who contends that the agency failed to
account for the impacts of automation on demand, capacity, and revenue assumptions.
These examples suggest that an effective response to automated vehicle technologies requires an early and
ongoing dialogue between regulators and developers.
As of late 2014, almost all regulatory action regarding automated cars and other on-road vehicles has
focused on the conditions in which on-road testing and operation of these vehicles takes place, including
vehicle and driver licensing (see annex). This is of course an essential element in the progression towards
the commercial deployment of increasingly automated and ultimately self-driving vehicles. Crucially,
however, driver and vehicle licensing are not the only regulatory elements that will be challenged by the
deployment of highly automated vehicles. As noted earlier, the urban mobility pathway leading to the
deployment of on-demand automated vehicle services is a plausible development in many urban areas.
In these cases, automated on-demand mobility systems will provide services similar to those provided by
taxis and public transport, two highly regulated industries. Authorities will have to adapt - and possibly
rethink - their approaches to regulating these activities in order to avoid conflicts. Failure to do so might
even prevent the deployment of the urban mobility pathway for autonomous vehicles which would stifle
innovative uses for this technology and potentially lead to welfare losses.
Regulatory considerations and policy choices
Treat automated vehicles specifically or generally?
A government seeking to expressly regulate automated driving could carefully examine and, as needed,
modify each existing law to clarify its application to vehicle automation. It could locate each regulatory
function in the agency ordinarily assigned to perform it. And it could seek, wherever possible, to apply
similar requirements to automated vehicles and their operators as would be applied to all others.
Alternately, such a government could promulgate a specific package of largely standalone rules that apply
exclusively to vehicle automation. It could give authority to implement these rules to a more limited set of
new or existing agencies. And it could intentionally differentiate between automated and non-automated
driving with respect to particular rights, obligations, and liabilities.
A general approach will likely seem more appropriate as automation becomes more widespread and
ordinary. Nonetheless, a specific approach, which more closely mirrors recent legislative and regulatory

output in the United States, may be a simpler and cleaner method of addressing a nascent set of
technologies that require development of particular expertise.
The paper “Automated vehicles are probably legal in the United States” (Smith, 2014c) contains model
language to clarify the legal status of automated driving that integrates existing vehicle codes wherever
possible and yet treats automated driving specifically when needed. For example, the language embraces
and expands existing registration obligations and standards rather than introducing wholly new ones.
Let policy lead or lag technology?
Proactive policy, including specific rules, can provide companies the legal clarity they need to make
investment and deployment decisions and can enable governments to appropriately handle automation
technologies at each of the regulatory gates described above.
Nonetheless, prematurely codifying requirements can freeze unrealistically high or low expectations into the
law in a way that ultimately causes that law to lag rather than to lead. Furthermore, duplicative or repeat
efforts to develop rules can force developers to invest resources in lengthy legislative debates and
regulatory processes.
For these reasons, informal dialogue may often be preferable to specific rules. Importantly, countries and
regions with a specific “automated driving law” are not necessarily ahead of those without one. The US
state of Michigan, for example, recently enacted a law that explicitly prohibits the operation of automated
vehicles for any purpose other than research and development testing. Where other US states may have
flexibility to accommodate new kinds of pilot projects that do not qualify strictly as testing, Michigan will
have none.
Privilege uniformity or flexibility?
Uniform regulation across multiple jurisdictions can reduce the cost and complexity for developers of
systems that will necessarily cross national or subnational borders. Designing for one set of rules and roads
is daunting enough; designing for dozens only amplifies this challenge. Flexible regulation, however, might
more easily accommodate existing regional differences, local start-ups, and unique demonstration projects.
It may also foster more national ownership over what is viewed by some as an area of international
competition.
Recently endorsed amendments to the 1968 Vienna Convention on Road Traffic strike an impressive
balance by delegating some relevant decisions to an international administrative structure. Another hybrid
approach, developed more below, would embrace significant regional flexibility for testing and pilot projects
but would emphasise consistency and reciprocity for production vehicles.
Emphasise ex-ante or ex-post regulation?
The choice between ex-ante regulation (particularly regulatory standards) and ex-post regulation
(particularly recalls and civil suits) also implicates flexibility. Forward-looking rules provide more certainty
but less flexibility; backward-looking measures provide more flexibility but less certainty. These trade-offs
are particularly relevant to concerns raised about the liability of automated vehicle manufacturers and
associated companies for injuries related to their products. These concerns, however, likely derive at least
as much from technical uncertainty (how will these eventual products actually perform) as from legal
uncertainty (how will courts determine liability)?

28 – REGULATORY ENVIRONMENT
Key challenges
Provide sufficient legal clarity
Governments should seek to provide sufficient legal clarity with respect to:
 Obligations of developers: Under what conditions can a developer test and market its vehicles?
 Obligations of operators: Who or what determines whether the operator of an automated vehicle
must be vigilant?
 Liabilities of developers: Are any ex-ante deviations from existing rules of criminal and civil liability
demonstrably justifiable?
 Liabilities of operators and owners: Are existing rules of civil liability sufficiently clear for companies
that provide motor vehicle insurance?
 Nonconventional operators: Is the operator of an automated vehicle sufficiently clear in existing
law, and are the requirements on that person sufficiently flexible to adjust based to emerging
technologies?
 Nonconventional motor vehicles: Is the legal status of low-speed passenger and delivery shuttles
sufficiently clear and appropriate?
 Nonconventional transport services: Is the legal status of ride sharing, car sharing, and non-
traditional taxicab services sufficiently clear and appropriate?
 Identify and correct market failures.1
Driving can impose costs on others - particularly through injuries and pollution - that are not internalised
by vehicle drivers or owners. Moreover, some of the costs that drivers and owners do bear, like vehicle
insurance, have an attenuated connection to the nature or extent of their driving. These mismatches can
change incentives in ways that encourage more driving.
Moreover, if automated vehicles actually do represent an improvement in safety or fuel efficiency, these
mismatches may also obscure financial benefit of these gains. For this reason, internalising more of the
costs of driving on an incremental basis (such as per kilometre travelled) may strengthen the individual
economic case for purchasing or using an automated vehicle.
Rationalise insurance
Expanding public insurance and facilitating greater private insurance could provide at least three benefits.
First, providing sufficient compensation to those injured by an automated vehicle could relieve some of the
pressure on the tort system to provide such a remedy. Second, enhanced vehicle insurance requirements,
especially if combined with greater flexibility in the administration of this insurance, could provide a third-
party check on the safety of automated systems. Third, such requirements could also help remedy the
potential market failures referenced above. Ultimately, however, as automation increases, liability could
gradually shift from drivers to manufacturers and OEMs although the allocation of liability among these
parties remains challenging and adjudication methods have yet to be developed.
Encourage information sharing by private developers
Education of public actors and of the public at large is essential to the development of effective regulations
and realistic expectations. Governments can facilitate this education by encouraging developers to share
1
This and the following challenges are adapted from (Smith, 2014d).

specific data about their products and processes. This can take at least two forms. First, companies that
request specific legal changes, particularly with respect to tort liability, should be expected to produce
information justifying these requests: If liability is an impediment, for example, the company should point
to specific products that it would release absent that liability. Second, if regulatory approval is required, a
key part of that approval could involve review of a specific safety case, preferably a public one, made by a
developer on behalf of its product. Pointing to particular processes for development, verification, and
monitoring, for example, could contribute to greater regulatory competence and greater public confidence.
Limit the duration of risk associated with early development stages
Vehicles with automated driving systems that are introduced in the next few years will be neither perfect
nor stay with us for only a few years. Years after they have become outdated, many of these vehicles will
still be on the road. A key goal for both regulators and developers should be limiting the physical risk of
these systems through a variety of technical and contractual tools to enable monitoring, over-the-air
updates, and even virtual recalls.
Expect as much from non-automated systems as from automated systems
Finally, automated driving has inspired much excitement and trepidation. Unfortunately, the shocking toll of
motor vehicle crashes today is often lost in these discussions. Regulators that impose heightened
requirements on automated vehicles should also expect more from the designers and drivers of non-
automated vehicles. Strategies to reduce intoxicated driving, distracted driving, speeding, poor vehicle
maintenance, and other common behaviours that substantially increase crash and injury risk can also
increase the attraction of automated driving.
Proposals for different automation strategies
Key measures for “something everywhere” automation
Because the “something everywhere” approach embraced by car manufacturers depends more on large
markets than on local conditions, sufficient consistency across jurisdictions in terms of physical, physical-
digital, digital, and legal infrastructure may facilitate smoother deployment. Moreover, ensuring that the
costs of driving are internalised, as noted above, can enable automated vehicles, if actually safer and more
fuel efficient, to claim an advantage over their non-automated counterparts.
Key measures for “everything somewhere” automation
In contrast to the “something everywhere” approach, the “everything somewhere” approach depends much
more on local conditions. For this reason, flexibility for individually tailored solutions and specialised pilot
projects are particularly important. Local, regional, and national actors can also facilitate this approach to
automation by clarifying the legal status of non-conventional vehicles (such as low-speed neighbourhood
shuttles) and services (for instance ride sharing and alternatives to traditional taxi service) by identifying
specific needs and opportunities, by modifying physical and digital infrastructures accordingly, and by
deploying particular resources such as road space or taxi concessions strategically.

30 – ANNEX
Annex: Current regulatory frameworks for the testing of
autonomous vehicles
The following overview reflects the state of regulation in early 2015.
United States
Four US states (California, the District of Columbia, Florida and Nevada) have passed laws allowing and
setting the conditions for the testing of automated and highly autonomous vehicles. Eleven states are
considering legislation addressing the testing of these vehicles and an equal number of states have failed to
pass bills allowing the on-road testing of autonomous vehicles.
There is, as of yet, no uniform approach to regulating autonomous vehicles among those states that have
passed legislation. All four states that have passed autonomous vehicle legislation allow non-testing use of
those vehicles, though in the case of Michigan, the driver/operator must be a representative of the
manufacturer. In California, Nevada and Florida, vehicles must meet Federal Motor Vehicle Safety
Standards. California, Nevada and the District of Columba require autonomous vehicles to have an easy to
trigger auto-drive disengage switch and an alert system for system failures. Both California and Nevada
require vehicles to store sensor data 30 seconds before a collision. Nevada restricts testing to specific
geographic contexts and California reserves the right to do so. Neither Florida nor the District of Columbia
impose geographic restrictions. Nevada only issues registration permits explicitly for testing whereas
Michigan only issues registration certificates to manufacturers.
More at http://www.aamva.org/WorkArea/linkit.aspx?LinkIdentifier=id&ItemID=5827&libID=5803 and
http://cyberlaw.stanford.edu/wiki/index.php/Automated_Driving:_Legislative_and_Regulatory_Action.
Europe
Jurisdictions where an ad hoc regulatory framework is in the making:
United Kingdom: As part of the 2013 National Infrastructure Plan, the UK government pledged a review of
the legislative and regulatory framework to enable trialling of driverless cars on UK roads. The review is
part of wider government action that includes a proposal to invest up to GBP 10 m in collaborative R&D
projects to research this area in the UK.
On 30 July 2014, the government announced a driverless cars competition, inviting UK cities to join
together with businesses and research organisations and host vehicle trials locally. On 4 August 2014, the
Department for Transport launched a public consultation concerning the review of the legislative and
regulatory framework for testing driverless cars. The scope of the review is limited to the testing of
driverless cars with high automation (as opposed to fully autonomous cars), that is to say, of cars which are
capable of operating on the road network without human intervention, but are fitted with a full set of
driving controls, and in which a driver must be able and ready to assume control. The consultation closed
on 19 September 2014. The review was published by the Secretary of State for Transport in December
2014. Relevant information and discussion document, available at:
https://www.gov.uk/government/consultations/driverless-cars-regulatory-testing-framework.
In February 2015, the UK Department for Transport released the results of this consultation in a set of
reports under the heading of “Driverless Cars in the UK: A Regulatory Review”. This review found that those
wishing to test automated and highly autonomous vehicles are not limited to test tracks or specific
geographic areas, nor are they required to obtain special certificates or permits. Further, they are not

ANNEX – 31
required to post surety bonds provided they have adequate insurance coverage. These reports can be found
at https://www.gov.uk/government/publications/driverless-cars-in-the-uk-a-regulatory-review.
France: The French Government announced the adoption of the necessary legal framework to enable the
testing of driverless cars on public roads as of 2015 under strict conditions (see press release no. 194,
2 July 2014, available at: http://proxy-pubminefi.diffusion.finances.gouv.fr/pub/document/18/17720.pdf.
The Netherlands: On 16 June 2014, the Dutch Minister of Infrastructure and Environment wrote a letter to
the Dutch House of Representatives, announcing that regulations are in the making to make large-scale
public road tests legal. Currently, Dutch drivers have limited abilities to hand control of their steering wheel
over to a car’s computer. Automatic parking for instance is permitted, but to make more extensive
manoeuvres possible regulation needs to change. The Minister plans to propose changes to the current
regulations by early 2015. Until then, only small scale tests are possible. Letter and relevant info available
at: http://www.techhive.com/article/2363920/large-scale-tests-with-self-driving-cars-to-hit-dutch-
roads.html.
Finland: In a press release of 21 May 2014, it was announced that the Ministry of Transport and
Communications is preparing an amendment to the Road Traffic Act that would allow for driverless robotic
cars to drive within a restricted area on public roads. The act in question would constitute experimental
legislation that would be in force for five years starting at the beginning of 2015. Robotic cars could be
tested, subject to a permit, in areas defined by the Finnish Transport Safety Agency. The testing of robotic
cars in public road traffic would be possible within limited time periods and in predetermined areas. Press
release available at: http://www.lvm.fi/pressreleases/4404066/mintc-to-launch-an-experiment-that-would-
allow-for-robotic-cars.
Jurisdictions without ad hoc regulatory framework, but where testing is permitted or has been permitted:
Sweden: In Sweden, there is no AV testing-specific regulatory framework. Yet, apparently, testing is
allowed. For instance, the Swedish city of Gothenburg has given Volvo permission to test 100 driverless
cars, as part of a two-year test scheme, called “Drive Me”, scheduled to start on 1 January 2017.The test
scheme will be the first in the world to put autonomous cars into the hands of ordinary drivers, in amongst
normal daily traffic, in significant numbers. Nevertheless, there seems to be no legal clarity at the moment
as to the conditions under which such testing will be carried out. Reportedly, a Volvo technical specialist
admitted that the company is not yet sure how Swedish law might need to be changed in order to enable
the public to take part in Drive Me and added that without legislative clarity the launch in 2017 will not be
able to go ahead (see: http://www.wired.co.uk/news/archive/2014-05/30/eu-embrace-self-driving-cars).
Germany: In Germany there is no specific legal framework on the testing of AVs, but testing in traffic is
allowed on the basis of a special permission. Several local and long-distance tests have been undertaken by
car manufacturers and original equipment manufacturers. Germany has also announced the creation of an
autonomous vehicle testing corridor on the A9 motorway between Berlin and Munich. This corridor will allow
testing of connected and autonomous vehicles and vehicle to infrastructure communication. The absence of
an ad hoc legal framework might be due to the strict interpretation of the Vienna Convention followed in
Germany. On this point, see: http://www.research-in-germany.org/en/research-landscape/news/news-
archive/2014/02/2014-02-18-who-should-be-legally-responsible-for-autonomous-cars-.html.
Spain: Testing of autonomous vehicles has also occurred in Spain, most notably in the context of the
SARTRE project, funded by the European Commission and aiming to develop strategies and technologies to
allow vehicle platoons to operate on normal public highways. The first tests of these technologies occurred
on Spanish motorways, see: http://www.sartre-project.eu.

32 – ANNEX
Singapore and Japan
Singapore: In August 2014, the Singaporean Ministry of Transport announced the launch of CARTS – the
Committee for Autonomous Road Transport for Singapore. CARTS’ stated goals include “studying and test-
bedding the various AV technologies in our environment, laying the legal, regulatory and liability framework
to govern the operations of AVs, and exploring industry development and business opportunities”. Relevant
information available at: http://thinkingcities.com/new-era-of-self-driving-cars-will-transform-cities and
http://vulcanpost.com/18369/driverless-vehicles-next-step-future-singapore.
On 27 August 2014, the Land Transport Authority (LTA) announced that it had signed a Memorandum of
Understanding (MOU) with the lead agency for R&D in Singapore for a joint partnership to set up the
Singapore Autonomous Vehicle Initiative (SAVI). The formation of SAVI will support CARTS to holistically
chart the strategic direction for AV-enabled land mobility concepts in Singapore. To support the R&D of AV
technology, LTA will work towards a framework that will allow the testing of AVs on the public road network
over 2015 (see: http://app.lta.gov.sg/apps/news/page.aspx?c=2&id=29525082-5265-4139-bc3b-
0241a4639d46).
Japan: Public road tests of AVs have been carried out also in Japan. In 2013, Nissan carried out Japan's
first public road test of an autonomous vehicle on a highway (relevant information available at:
http://www.bbc.com/news/technology-28551069). Reportedly, self-driving trucks have also been tested in
Japan (see: http://www.theverge.com/2013/2/27/4037568/self-driving-trucks-tested-in-japan).

BIBLIOGRAPHY – 33
Bibliography
CityMobil2, 2015. CityMobil2: Cities demonstrating automated road passenger transport. [Online]
Available at: http://www.citymobil2.eu/
[Accessed 10 April 2015].
NHTSA, 2013. U.S. Department of Transportation Releases Policy on Automated Vehicle Development.
[Online]
Available at:
http://www.nhtsa.gov/About+NHTSA/Press+Releases/U.S.+Department+of+Transportation+Releases+Poli
cy+on+Automated+Vehicle+Development
[Accessed 15 October 2014].
Parrish, K., 2015. http://schedule.sxsw.com/2015/events/event_IAP37302. [Online]
Available at: http://schedule.sxsw.com/2015/events/event_IAP37302
[Accessed 10 April 2015].
SAE, 2014. Taxonomy and Definitions for Terms Related to On-Road Motor Vehicle Automated Driving
Systems. [Online]
Available at: http://standards.sae.org/j3016_201401/
Smith, B. W., 2013. Testing, certification, and licensing. [Online]
Available at: http://bit.ly/1dUIvvn
Smith, B. W., 2014a. A Legal Perspective on Three Misconceptions in Vehicle Automation. In: G. a. B. S.
Meyer, ed. Road Vehicle Automation Lecture Notes in Mobility. s.l.:Springer.
Smith, B. W., 2014b. Proximity-Driven Liability. Georgetown Law Journal, Issue 102.
Smith, B. W., 2014c. Automated Vehicles Are Probably Legal in the United States. Texas A&M Law Journal,
Volume 1.
Smith, B. W., 2014d. Regulation and the Risk of Inaction. [Online]
Available at: http://www.aamva.org/Autonomous-Vehicle-Information-Library/

2015-05/Photocredit:istockphoto.com/mathisworks
International Transport Forum
2 rue André Pascal
75775 Paris Cedex 16
France
T +33 (0)1 45 24 97 10
F +33 (0)1 45 24 13 22
Email : itf.contact@oecd.org
Web: www.internationaltransportforum.org
Automated and Autonomous Driving
Many cars sold today are already capable of some level of automated
operation, and prototype cars capable of driving autonomously have
been and continue to be tested on public roads in Europe, Japan and the
United States. These technologies have arrived rapidly on the market and
their future deployment is expected to accelerate. Autonomous driving
promises many benefits: improved safety, reduced congestion and lower
stress for car occupants, among others.
But authorities will have to adapt existing rules and create new ones in
order to ensure the full compatibility of these vehicles with the public’s
expectations regarding safety, legal responsibility and privacy. This
report explores the strategic issues that will have to be considered by
authorities as more fully automated, and ultimately autonomous, vehicles
arrive on our streets and roads. The report was drafted on the basis of
expert input and discussions amongst project partners in addition to a
review of relevant published research and position papers.
The work for this report was carried out in the context of a project
initiated and funded by the International Transport Forum’s Corporate
Partnership Board (CPB). CPB projects are designed to enrich policy
discussion with a business perspective. Led by the ITF, work is carried
out in a collaborative fashion in working groups consisting of CPB
member companies, external experts and ITF researchers.

Automated & Autonomous Driving

More Related Content

What's hot (20)

Viewers also liked (15)

Similar to Automated & Autonomous Driving (20)

More from Roger Atkins (20)

Recently uploaded (20)

Automated & Autonomous Driving