Scenarios about development and implications of automated
1vehicles in the Netherlands
23 4
Dimitris Milakis* 5
Assistant Professor, Department of Transport and Planning,
6
Faculty of Civil Engineering and Geosciences, Delft University of Technology
7 Tel. +31 15 27 84981 8 E-mail: d.milakis@tudelft.nl 9 10 Maaike Snelder 11
Researcher, Department of Transport and Planning,
12
Delft University of Technology;
13
TNO Netherlands Organization for Applied Scientific Research
14 Tel. +31 15 27 84981 15 E-mail: M.Snelder@tudelft.nl 16 17
Bart van Arem 18
Professor, Department of Transport and Planning,
19
Faculty of Civil Engineering and Geosciences, Delft University of Technology
20 Tel. +31 15 27 86342 21 E-mail: b.vanarem@tudelft.nl 22 23
Bert van Wee 24
Professor, Transport and Logistics Group,
25
Faculty of Technology, Policy and Management, Delft University of Technology
26 Tel. +31 15 27 81144 27 E-mail: G.P.vanWee@tudelft.nl 28 29
Gonçalo Homem de Almeida Correia 30
Assistant Professor, Department of Transport and Planning,
31
Faculty of Civil Engineering and Geosciences, Delft University of Technology
32 Tel. +31 15 27 81384 33 E-mail: G.Correia@tudelft.nl 34 35 36 37 38 39 40
6235 Words + 5 Figures + 2 Tables = 7985 Words
41 42
Submitted for presentation to the 95th Annual Meeting of the Transportation Research Board 43 44 *Corresponding author 45 46 47 48 49 50
ABSTRACT 1
2
Automated driving technology is emerging. Yet, less is known about when automated
3
vehicles will hit the market, how penetration rates will evolve and to what extent this
4
new transportation technology will affect transportation demand and planning. This
5
study identified through scenario analysis plausible future development paths of
6
automated vehicles in the Netherlands and estimated potential implications for traffic,
7
travel behavior and transport planning on a time horizon up to 2030 and 2050. The
8
scenario analysis was performed through a series of three workshops engaging a
9
group of diverse experts. Sixteen key factors and five driving forces behind them were
10
identified as critical in determining future development of automated vehicles in the
11
Netherlands. Four scenarios were constructed assuming combinations of high or low
12
technological development and restrictive or supportive policies for automated
13
vehicles (AV …in standby, AV …in bloom, AV …in demand, AV …in doubt).
14
According to the scenarios, fully automated vehicles are expected to be commercially
15
available between 2025 and 2045, and to penetrate market rapidly after their
16
introduction. Penetration rates are expected to vary among different scenarios
17
between 1% and 11% (mainly conditionally automated vehicles) in 2030 and between
18
7% and 61% (mainly fully automated vehicles) in 2050. Complexity of the urban
19
environment and unexpected incidents may influence development path of automated
20
vehicles. Certain implications on mobility are expected in all scenarios, although there
21
is great variation in the impacts among the scenarios. It is expected that measures to
22
curb growth of travel and subsequent externalities will be necessary in three out of the
23
four scenarios.
24 25
Keywords: Automated vehicles, scenarios, development, implications, The 26 Netherlands 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
INTRODUCTION 1
2
The introduction to the market, the development and the implications of automated
3
driving are among the main uncertainties of the future transport system. According to
4
Milakis et al. (1) automated vehicles could have significant impacts on cities and
5
transportation systems. The design of robust long-term transport policies and
6
investments needs to take into account uncertainties associated with automated
7
vehicles. The aim of this study is to identify plausible future development paths of
8
automated vehicles in the Netherlands and to estimate potential implications for
9
traffic, travel behavior and transport planning on a time horizon up to 2030 and 2050.
10
The research questions are the following:
11
o What are the possible developments for automated vehicles and which factors will
12
determine these developments on a time horizon up to 2030 and 2050 in the
13
Netherlands? What stages in this development can be distinguished?
14
o What are the implications for road capacity? Does this differ between urban roads,
15
regional roads and motorways?
16
o What are the implications for users (value of time) and consequently for travel
17
behavior?
18
o To what extent might automated vehicles affect transportation planning?
19
The rest of this paper is structured as follows. The literature review is firstly
20
presented. Then we describe our methodology and we present four scenarios about the
21
development and possible effects of automated vehicles in the Netherlands. We close
22
this paper with our conclusions.
23 24
LITERATURE REVIEW 25
26
Development of automated vehicles 27
28
The development of automated vehicles takes place along two axes. First, the manual
29
to automated axis that describes the extent to which the human driver monitors the
30
driving environment and executes aspects of the dynamic driving task. Second, the
31
autonomous to cooperative axis that describes the extent to which vehicles can
32
communicate and exchange information with other vehicles (V2V) or with the
33
infrastructure (V2I). For the manual-to-automated axis, two main taxonomies that
34
classify the levels of vehicle automation have been identified (2, 3). In both
35
taxonomies the first three levels assume that a human driver will control all driving
36
tasks. The remaining levels assume that an automated driving system takes control of
37
all dynamic tasks of driving. In our scenario study we use the NHTSA (2) taxonomy
38
for vehicle automation. We refer to level 3 and level 4 as ‘conditional’ and ‘full’
39
automation respectively. In conditional automation the driver is expected to be
40
available for occasional control of the vehicle while in full automation s/he is not. Full
41
automation comprises both occupied and unoccupied vehicles. According to a survey
42
held during the Automated Vehicles Symposium 2014, experts expect conditionally,
43
and fully automated vehicles to hit the market in 2019 and 2030 respectively (median
44
values) (4). Litman (5) estimated the penetration rate of fully automated vehicles
45
assuming that they will hit the market in 2020. He based his estimations on the
46
deployment of previous vehicle technologies like air bags, automatic transmission,
47
navigation systems, GPS services, hybrid vehicles and on assumptions about the
48
purchase price of these vehicles. He concluded that, in the United States, it may take
49
ten to thirty years from the time of launch before the automated vehicles dominate the
car sales market and another ten to twenty years before the majority of travel is done
1
using automated vehicles. Moreover, Kyriakidis et al. (6) reported that 69% of the
2
respondents in their internet-based questionnaire survey expected fully automated
3
vehicles to reach 50% penetration rate up to 2050.
4 5
Impacts of automated vehicles on road capacity, travel behavior and transport 6
infrastructures 7
8
Automated vehicles could enhance road capacity by optimizing driving behavior with
9
respect to time gaps, speed and lane changes (7). The magnitude of this impact is
10
related to the level of automation and cooperation between vehicles. Thus far,
11
literature focuses on the automation of longitudinal driving, with the help of adaptive
12
cruise control (ACC) and cooperative adaptive cruise control (CACC). These studies
13
indicate that ACC can either have a small negative or a small positive effect on
14
capacity (-5% to +10%) (see e.g. 8, 9). For CACC, most studies report a quadratic
15
increase in capacity as the penetration rate increases, with a theoretical maximum
16
increase of 100% (doubling). These studies indicate that the increase in capacity is
17
high (>10%) only if the penetration rate is higher than 40% (see e.g. 10, 11, 12).
18
The effect of automated vehicles on travel behavior is a relatively less
19
researched topic. Malokin et al. (13) showed that the ability of multitasking in
20
automated vehicles could increase driving alone and shared ride commute shares by
21
1% each. Gucwa (14) applied several scenarios of potential changes in value of time
22
because of introduction of automated vehicles in San Francisco. He found an increase
23
in vehicle kilometers traveled (VKT) between 4% and 8% for different scenarios.
24
Moreover, two simulation studies reported increased VKT rates for automated
25
(vehicle and ride) sharing schemes compared to privately owned conventional
26
vehicles (15, 16).
27
The introduction of automated vehicles could reduce the requirements for road
28
network expansion in the future because of the increases in road capacity.
29
Estimations about the magnitude of this reduction vary from substantial (17) to only
30
marginal because of the possibility of induced travel demand (5, 18, 19). Automated
31
vehicles could also challenge the role of public transport (buses in particular) in the
32
future transport system (5, 20). Finally, increases in road capacity may create the
33
opportunity for development of bicycle and pedestrian infrastructures (i.e, bicycle
34
lanes or wider sidewalks).
35 36
METHODS 37
38
Given the uncertainty and the long-term character of our research questions, a
39
scenario analysis was applied. Scenario analysis is “the process of evaluation possible
40
future events through the consideration of alternative plausible, though, not equally 41
likely, states of the world” (21). Scenarios have the advantage over forecasts in that 42
they are more flexible, creative and not necessarily probabilistic outlines of plausible
43
futures. Thus, they could assist long-term planning to broaden perspectives and
44
identify key dynamics (21). According to Maack (22) and Townsend (23) scenarios
45
should be plausible, distinctive, consistent, relevant, creative, and challenging.
46
Whilst methodologies for scenario construction show great variation, there is a
47
basic common underlying structure (see 22, 24, 25). In our study, the scenario
48
development process involved five sequential steps: (a) identification of key factors
49
and driving forces of development of automated vehicles, (b) assessment of impact
and uncertainty of driving forces, (c) construction of the scenario matrix, (d)
1
estimation of penetration rates and potential implications of automated vehicles in
2
each scenario, and (e) review of the scenarios and assessment of the likelihood and
3
overall impact of each scenario.
4
The process was completed in three workshops. The first two workshops
5
involved five experts (the authors of this paper) from Delft University of Technology.
6
In the first workshop, we identified the key factors of development of automated
7
vehicles in the Netherlands and the driving forces behind them. Each expert was also
8
asked to rank the driving forces with respect to the magnitude of their potential effect
9
on development of automated vehicles (impact) and the predictability of their future
10
state (uncertainty). A scenario matrix was subsequently drafted based on the results of
11
the first workshop. In the second workshop, we discussed penetration rates of
12
automated vehicles in 2030 and 2050 in the Netherlands, and potential implications
13
for road capacity, value of time and VKT in each of the four scenarios. After the
14
workshop, a questionnaire was distributed to the participants asking numerical
15
estimations about all issues discussed in the second workshop.
16
In the last workshop, the four draft scenarios were presented and reviewed by
17
fifteen experts from planning, technology, and research organizations in the
18
Netherlands (e.g., I&M - Ministry of Infrastructure and the Environment, RWS -
19
Ministry of Transport, Public Works, and Water Management, Connekt, KiM -
20
Netherlands Institute for Transport Policy Analysis, RDW - National road traffic
21
agency, Spring Innovation, Eindhoven University of Technology). The discussion was
22
organized in two sessions ((a) development and (b) implications of automated
23
vehicles in the Netherlands) and was coordinated by the five experts from Delft
24
University of Technology. All twenty experts also evaluated the scenarios in terms of
25
likelihood and overall impact (i.e., value of time, road capacity, and total VKT).
26 27 RESULTS 28 29
Key factors and driving forces 30
31
Sixteen key factors and five driving forces behind them (policies, technology,
32
customers’ attitude, economy and the environment) were identified as critical in
33
determining future development of automated vehicles in the Netherlands (see Table
34
1). The driving forces were assessed with respect to the magnitude of their potential
35
effect on development of automated vehicles (impact) and the predictability of their
36
future state (uncertainty). According to the results, technology is expected to have the
37
strongest impact on the development path of automated vehicles, but it is also highly
38
unpredictable (see Table 2). Policies were found to be quite influential, but uncertain
39
as well. Customers’ attitude was also indicated as a highly unpredictable factor, but
40
the expected impact was assumed to be lower than technology and policies. Finally,
41
economy and the environment were assumed to be fairly predictable and to have
42
relatively lower impact on the development of automated vehicles.
43
Based on those results, technology and policies appeared to be the most
44
influential driving forces. Both were also highly unpredictable although customers’
45
attitude appeared as equally uncertain driving force. Therefore, technology and
46
policies were selected as the most relevant driving forces to build our scenario matrix.
47 48 49 50
TABLE 1 Key Factors and Drivers of the Development of Automated Vehicles 1
in the Netherlands. 2
3
Key factors Driving forces
AV technology trials Technology, Policies Interoperability among AV technologies Technology, Policies
Costs/benefits of AV technology Technology, Policies, Customers’ attitude Development of AV in EU Technology, Policies, Customers’ attitude AV ownership structure (public vs private) Technology, Economy
Transition steps Technology, Policies
Incidences Technology
Energy, emissions Technology, Policies, Economy, Environment, Legal/institutional context (national and
European) Policies
Public/private expenditures on infrastructure Policies, Economy
Stability of policies Policies
Accessibility, social equity Technology, Policies Psychological barriers (Citizens and
customers)
Technology, Customers’ attitude Marketing/image of AV Policies, Customers’ attitude,
Attitudes towards AV Technology, Policies, Customers’ attitude, Economy, Environment
Income Economy
4
TABLE 2 Ranking (Median Values) of Driving Forces of the Development of 5
Automated Vehicles (AV) According to their Impact and Uncertainty (1-Lowest, 6
5-Highest). 7
8
Driving forces Impact Uncertainty
Technology 5.0 4.0 Policies 4.0 4.0 Customers’ attitudes 3.0 4.0 Economy 2.0 2.0 Environment 1.0 1.0 9 10 Scenario matrix 11 12
Four scenarios were constructed assuming combinations of high or low technological
13
development and restrictive or supportive policies for automated vehicles (AV …in
14
standby, AV …in bloom, AV …in demand, AV …in doubt; see Figure 1). Although
15
scenarios were built around permutations of those two driving forces, we also
16
incorporated all remaining driving forces in our scenario plots (customers’ attitude,
17
economy, environment). The aim was to capture as much as possible of the
18
complexity surrounding this exercise. Moreover, the key factors offered input into the
19
development of detailed, dynamic and coherent storylines. The scenario plots are
20
presented in the next four sections.
21 22 23 24
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
FIGURE 1: Scenario matrix about development of automated vehicles in the 23
Netherlands. 24
25
Scenario 1: Automated vehicles …in stand by 26
27
Although discussion about the potential of having fully automated vehicles on public
28
roads by 2030 had been intensified already since 2015 and conditional automation
29
was a reality since 2020, the Dutch government decided not to heavily invest on
30
integration of this mobility technology in the transportation system of the
31
Netherlands. In fact the government did not see any major benefits stemming from a
32
rapid development of automated vehicles, while they did foresee a lot of risks
33
associated with this technology. It is true that the Dutch transportation system was
34
really efficient in the early 2020’s with a multimodal character, which was translated
35
into a modal split where almost half of the trips were being undertaken by bicycle,
36
foot or public transportation. Moreover, transport safety was steadily improving and
37
no major environmental problems were expected in following years. The consistent
38
strategy towards low-carbon economy had already started paying-off. Also, the
39
modest economic growth did not allow allocation of more resources on infrastructures
40
related to this emerging technology (V2I).
41
The combination of Dutch government’s skepticism about automated vehicles
42
and the weak income growth had possibly played a role on customers’ moderate to
43
low demand for automated vehicles. Customers’ demand did not significantly change
44
even when conditionally automated vehicles were made commercially available in
45
2020. The fact that the Dutch government allowed conditionally automated vehicles
46
to travel only in motorways until 2025 might have deterred demand. Moreover, the
47
attitude towards vehicles in general and automated vehicles in particular was not very
48
positive at that time, with most customers adopting a ‘wait and see’ position. In fact
49
vehicles use had already reached its peak a decade earlier (during 2010’s) mainly
50
because of the generation Y reluctance to live an automobile oriented 20th century
51
like life. This attitude did not change dramatically in the following years until the
advent of fully automated cars in 2030. At that point automated vehicles (conditional
1
automation) represented a small fraction of total vehicles fleet (4%) and a slightly
2
higher percentage (7%) of total VKT (see Figure 2).
3
The advent of fully automated vehicles in 2030 signalized a change in
4
customers’ attitude. Auto manufacturers adopted an aggressive promotion strategy,
5
which among other actions allowed everyone to experience first hand a fully
6
automated vehicle for a week. They knew that ‘hands on’ experiences could remove
7
psychological barriers of automated driving from both customers and citizens, even if
8
eventually they would not buy the car. Moreover, seamless communication between
9
automated vehicles (V2V) and safe operation in urban environments signaled a huge
10
progress. Operation first of conditional and then of fully automated vehicles in urban
11
environments was indeed proven a real challenge especially with respect to urban
12
intersections and uncontrolled pedestrian movements. Customers’ attitudes about
13
automated vehicles became progressively more positive after 2030, which was
14
translated into stronger demand for this kind of vehicles. Twenty years later (2050)
15
automated vehicles represented 26% of vehicles fleet and 33% of total VKT (see
16
Figure 2). During the same period (2030-2050) the Dutch government regulated
17
several areas related to fully automated vehicles (e.g., automated-taxis, liability,
18
safety). However, no proactive actions were taken to further promote this mobility
19
technology because initial fears about potential negative implications, like strong
20
induced travel demand, sprawling trends and a modal shift from conventional public
21
transportation to automated vehicles were confirmed. In fact, the decrease of value of
22
time for automated vehicles users by 21% (see Figure 3) and the increase of
23
motorways capacity by 7% (mainly because of the development of cooperative
24
systems) (see Figure 4) could easily explain the increase of total VKT by 7% in 2050
25
(see Figure 5).
26 27
Scenario 2: Automated vehicles …in bloom 28
29
The CEO of Audi predicted in his interview on Automotive News in 2015 that “a
30
vehicle capable of driving itself with no need for any interaction from the driver, even 31
in critical situations, is probably 10 years away”. He was right. Technological 32
development between 2015 and 2025 was really rapid. First vehicles with conditional
33
automation were already launched in the market in 2018 and fully automated vehicles
34
hit the market in 2025. Governments in the Netherlands, Sweden, UK, Japan and
35
USA helped research community, high technology industry and auto manufacturers to
36
rapidly push the boundaries of vehicle cooperation (V2V) and automation. In the
37
Dutch context, a progressive regulatory framework for automated vehicles trials was
38
adopted as early as 2016, while significant investments in research and development
39
followed in coming years, supported by R&D funds of the European Commission.
40
Important investments on infrastructure communication with vehicles (V2I) were
41
decided in mid-2010’s and implemented within the next ten years, allowing for
42
seamless operation of automated vehicles in motorways and urban streets but also for
43
easy system upgrades thereafter. Moreover, an aggressive subsidy policy was
44
adopted. During first five years after launch, fully automated vehicles were exempted
45
from the registration fee, while electric automated vehicles were exempted from road
46
taxes as well. In the case of shared electric automated vehicles (automated taxis) the
47
government decided to provide an additional subsidy of 3000€ on the purchase, which
48
had been proved a successful measure for electric-taxis about a decade earlier. It was
49
clear that the Dutch government was seeing automated vehicles as the solution to
many long-standing mobility-related societal problems originated in 20th century, like
1
congestion and traffic fatalities. They were also considering the introduction of
2
automated vehicles as an opportunity for developing a more efficient multimodal
3
transportation system. The healthy macro-economic environment in Europe and the
4
high economic growth in the Netherlands supported the decisions for adopting such
5
aggressive promotional policies for automated vehicles. Moreover, most policy
6
reports from governmental organizations at that time were suggesting that investments
7
on automated vehicles were highly likely to pay-off soon by addressing many of the
8
inefficiencies of the conventional transportation system. An important prerequisite, as
9
all reports clearly noted, was user acceptance.
10
Customers’ attitude about automated vehicles evolved quite positively during
11
the 2010’s. It was the disruptive change in the mobility experience that attracted the
12
attention of most people at that point. More productive use of travel time and safe
13
driving conditions were among the changes that customers valued more. They were
14
also frequently referring to wider positive societal implications such as lower energy
15
consumption, environmental protection, economic, and social equity benefits (e.g.,
16
mobility for elderly and disabled persons). The positive economic context, the
17
supportive governmental policies, but also the wider societal changes of that period
18
such as the growth of digital and shared economy and the environmental awareness
19
movement played also a key role in having strong demand for automated vehicles.
20
The share of automated vehicles reached 11% in 2030 and rocketed to 61% in 2050
21
(see Figure 2). The share of VKT by automated vehicles in total travel followed a
22
similar path (23% in 2030 and 71% in 2050). As expected, automated vehicles users
23
(especially early adopters) were inclined to drive, on average, more kilometers than
24
users of conventional vehicles because of the opportunity they had to relax or do other
25
useful things during their trip. Indeed the value of time for automated vehicles users
26
had dropped 18% already by 2030 and 31% by 2050 (see Figure 3). New models of
27
fully automated vehicles after 2030 offered a highly flexible interior design that
28
allowed all kind of activities to be undertaken during travel including sleeping,
29
working, tele-conferencing and many more. Moreover, the combination of automated
30
and cooperative systems (V2V and V2I) allowed capacity to increase on both
31
motorways and urban streets by 25% and 6% respectively in 2050 (see Figure 4). All
32
these benefits did not come without a cost. Total vehicle kilometers had significantly
33
increased by 3% already in 2030 and by 27% in 2050 (see Figure 5). The Dutch
34
government quickly realized that congestion relief would not come simply by
35
introducing automated vehicles. In fact, they realized that congestion could get worse
36
in the future because of induced travel demand and sprawling trends, if they would
37
not take action. Therefore, stricter land use policies inspired by the compact city
38
paradigm (which had been abandoned decades earlier) and transportation demand
39
policies, such as road pricing, had been introduced during the 2040s to curb growth in
40
travel and urban expansion. Furthermore, automated taxis had been highly regulated
41
after 2030 with respect to total number of taxis per capita, and hours of operation.
42
Automated taxis were responsible for a significant part of VKT increase and thus
43
congestion, mainly because of their 24/7 non-stop operation. Dynamic policy
44
adaptation, such as in the case of automated taxis (from heavily subsidized to highly
45
regulated), was clearly the right way to go in a new transportation ecosystem where
46
asymmetric changes were more likely than ever.
47 48 49 50
Scenario 3: Automated vehicles …in demand 1
2
The optimism for seamless mobility by fully automated vehicles in the near future
3
was high in the mid 2010’s. The countless discussions in popular media were centered
4
on possible changes that this technology could bring to daily mobility and
5
subsequently to our societies. These discussions were fueled by frequent
6
announcements of auto manufacturers’ plans for fully automated mobility until 2025.
7
Many governments around the world, including the Netherlands, were foreseeing
8
major societal benefits from this technology, like congestion relief and significant
9
reduction of accidents. Therefore, they rapidly formed progressive legislative
10
frameworks allowing automated vehicles trials and supporting cooperation between
11
automobile and high tech industry. They also invested on research and development
12
of this technology and asked governmental organizations to adapt their plans to
13
possible development of vehicle automation in coming years. Moreover, they secured
14
important resources to fund smart infrastructures that would allow communication
15
with automated vehicles both on motorways and in urban environments (V2I). These
16
investments were partly funded by European Commission R&D funds, which in the
17
meantime had decided to allocate more resources in developing vehicle automation in
18
Europe mainly because of the expected traffic safety benefits.
19
However, the technological path to full automation was proved more difficult
20
than assumed. It took ten years (2025) for auto manufacturers in collaboration with
21
high technology companies only to make conditional automation commercially
22
available. The variability of road infrastructure and weather conditions, but also the
23
complexity of urban environment especially with respect to interaction with other
24
road users (conventional cars, cyclists and pedestrians) and to unexpected events (e.g.,
25
road flooding) required exhaustive tests and continuous adaptation of technology to
26
meet high safety standards. Moreover, the first fatal accidents in European urban
27
roads between conditionally automated vehicles and pedestrians in 2026 proved that
28
this technology was not entirely ready (at least for urban environments). The
29
European Union and many governments around the world responded with a mandate
30
that conditionally automated vehicles were only allowed on motorways until the
31
technology would evolve enough according to even higher safety standards. The
32
Dutch government also announced a new round of funding for research and
33
development in this area. Fifteen years later (2040) fully automated vehicles were
34
hitting the market.
35
Customer’s demand for automated vehicles incrementally increased up to
36
2040 and significantly expanded thereafter. Only 3% of total vehicle fleet was
37
(conditionally) automated in 2030 representing 5% of total VKT (see Figure 2). The
38
first fatal accidents in 2026 further prevented customers from buying automated
39
vehicles. It was only in 2040 with the advent of fully automated vehicles when the
40
psychological barriers for this technology were truly removed and sales subsequently
41
increased. In coming years people realized that this was a safe technology with
42
significant benefits especially with respect to comfort and to various activities
43
someone could undertake during a trip. The value of time was decreased by 16% for
44
automated vehicle users in 2050 (see Figure 3), while penetration was quite high at
45
that time with 17% of all vehicles being automated (see Figure 2). Moreover capacity
46
increased by 5% in motorways and by 2% in urban streets in 2050 (see Figure 4). The
47
combination of a decrease in value of time and an increase in capacity resulted in
48
more VKT in 2050 (3%) (see Figure 5). In fact the Dutch government was expecting a
49
stronger increase of VKT in coming years after 2050, because penetration of
automated vehicles was expected to become even higher. Therefore, they were
1
already planning to introduce travel demand management measures from the
2
beginning of 2050s to prevent major increase of VKT. Unfortunately, a significant
3
portion of automated vehicles was still carrying internal combustion engines. Thus,
4
increased VKT were associated with more energy consumption and more emissions.
5 6
Scenario 4: Automated vehicles …in doubt 7
8
Automated vehicles were one of the most appealing concepts of mobility technology
9
during the 20th century. No accidents, no driving effort, more personal time, less
10
congestion and almost no parking problems were the basic elements of vehicle
11
automation imprinted in the collective imaginary. In the early 21st century, the
12
discussion about the prospects of a fully automated mobility world resurfaced because
13
of some technological progress of auto manufacturers and high technology companies
14
in this area. However, full automation was still way too far from reality and cities and
15
transportation systems were more complex than ever. Thus, such a socio-technical
16
transition seemed quite difficult even if the technology was available.
17
In fact, it turned out that none of the basic forces (high technological
18
development, supportive policies and positive customers’ attitudes) for such a
19
transition were existent. In a recessive global economic context during late 2010’s,
20
most governments (the Dutch included) did not intend to spend their valuable
21
resources on research and on infrastructures for automated vehicles. Neither did they
22
develop a supportive institutional framework for testing and developing this
23
technology. They thought that vehicle automation might in fact lead to
counter-24
effective results for the transportation system. Their deepest fear was that the system
25
would not evolve enough to become fully automated. Thus, it was likely to stuck in
26
transition where semi-automated, fully automated and conventional cars would
co-27
exist, causing major safety and congestion problems especially in urban
28
environments. Their fears were absolutely justified. The technology evolved quite
29
slowly after 2015 with first conditional automation vehicles hitting the market only in
30
2028 and subsequently allowed to travel only on Dutch motorways. The bankruptcy
31
of a major automotive company (due to a sharp decrease in sales of conventional cars
32
in China) and the shift of attention of a high tech giant from automated vehicles to
33
other emerging technologies could partly explain the slow technological development
34
in this field. Technical difficulties associated with the detection of obstacles and
35
navigation in various road and weather conditions and in complex urban
36
environments inhibited rapid technological development as well.
37
As a result only 1% of total vehicles fleet was (conditionally) automated in
38
2030. Customers’ were reluctant to buy this technology since neither the government
39
supported it through, for example, subsidizing policies, nor middle-class income
40
could afford to pay for such a premium technology. When fully automated vehicles
41
were launched in 2045, customers’ interest became stronger since the benefits where
42
clearer then and the Dutch government allowed these vehicles to travel in urban
43
environments as well. However, the price for this technology was still too high, thus
44
fully automated vehicles continued to represent a marginal share of the vehicles’ fleet
45
in 2050 (7%). Moreover fully automated taxis offering premium services became
46
available after 2045. These companies have invested in transforming the interior of
47
these taxis into fully functional work and rest spaces. The marginal share of
48
automated vehicles affected neither capacity nor total VKT in 2050.
Unlike 20th century, vehicle automation did make it through to the market in
1
21st century. However, until 2050 automated vehicles was still a technology for the
2
upper class that could afford to have it. The rest of the people could have an
3
experience with automated vehicle by hiring an automated taxi or by just taking
4
automated buses, which in the meantime had grown rapidly.
5 6
7
8
9
10
11
12
13
14
FIGURE 2 Estimation of (a) percentage of (conditionally and fully) automated 15
vehicles in vehicles fleet and (b) percentage of VKT by (conditionally and fully) 16
automated vehicles in total VKT, in 2030 and 2050. Each bar represents average 17
value of five experts responses collected in Delft University of Technology 18
workshops and error bar depicts standard deviation. 19 20 21
22
23
24
25
26
27
28
FIGURE 3 Estimation of decrease in value of time of automated vehicle users in 29
different scenarios. Each bar represents average value of five experts responses 30
collected in Delft University of Technology workshops and error bar depicts 31
standard deviation. 32
1
2
3
4
5
6
7
8
9
10
11
12
13 14
FIGURE 4 Estimation of capacity changes in different scenarios. Each bar 15
represents average value of five experts responses collected in Delft University of 16
Technology workshops and error bar depicts standard deviation. 17
18
19
20
21
22
23
24
FIGURE 5 Estimation of change in total vehicle kilometres traveled in different 25
scenarios. Each bar represents average value of five experts responses collected 26
in Delft University of Technology workshops and error bar depicts standard 27
deviation. 28
Likelihood and overall impact of the scenarios 1
2
All scenarios were assessed with respect to their likelihood and overall impact during
3
the last workshop. The participants were asked to evaluate each scenario with respect
4
to (a) its likelihood, on a scale ranging from 0% (impossible) to 100% (certain), and
5
(b) its overall impact (i.e., on value of time, road capacity, and total VKT) on a scale
6
ranging from 0 (no impact) to 5 (highest impact). The participants were also asked to
7
respond about their confidence with the estimations on a scale ranging from 0 (not at
8
all confident) to 5 (very confident).
9
According to the results, scenario 2 (AV ….in bloom) and scenario 3 (AV
10
…in demand) were perceived as the most likely to happen in the future (41.8% and
11
38.3% respectively). The likelihood of scenario 1 (AV …in standby) and of scenario
12
4 (AV …in doubt) was assessed lower (31.8% and 25.8% respectively). The average
13
sum of probabilities per person was 137.5%, while only one person estimated a sum
14
of probabilities below 100%. Moreover, the participants were quite confident about
15
their responses (average level of confidence: 3.1). The results did not change
16
significantly when responses were weighted based on the level of confidence.
17
The scenario 2 (AV …in bloom) was also expected to have the highest overall
18
impact (4.6), with scenario 1 (AV …in standby) and scenario 3 (AV …in demand)
19
having similar but lower effects (2.4 and 2.3 respectively). The scenario 4 (AV …in
20
doubt) was not expected to have significant impacts on mobility (1.1).
21 22
CONCLUSIONS 23
24
The aim of this study was to identify plausible future development paths of automated
25
vehicles in the Netherlands and to estimate potential implications for traffic, travel
26
behavior and transport planning on a time horizon up to 2030 and 2050. To this end a
27
scenario analysis was conducted. Technology and policies were assessed to be the
28
most influential and unpredictable driving forces; hence the scenario matrix was built
29
around them. Four scenarios were constructed assuming combinations of high or low
30
technological development and restrictive or supportive policies for automated
31
vehicles. All remaining driving forces (customers’ attitude, economy and the
32
environment) have been incorporated in our scenarios as well.
33
According to our scenario analysis:
34
o fully automated vehicles are expected to be commercially available within a time
35
window of twenty years (between 2025 and 2045), while the respective
time-36
window for conditional automation is smaller (ten years) and more immediate
37
(between 2018 and 2028),
38
o full vehicle automation will likely be a game changer, driving the demand for
39
automated vehicles high. Penetration rates of automated vehicles are expected to
40
vary among different scenarios between 1% and 11% (mainly conditionally
41
automated vehicles) in 2030 and between 7% and 61% (mainly fully automated
42
vehicles) in 2050,
43
o vehicle automation and cooperation will likely follow converging evolution paths.
44
The type of cooperation (V2I, V2V) will likely vary though among different
45
scenarios according to the main drivers (policies, technological development),
46
o complexity of the urban environment is expected to influence the development
47
path of automated vehicles either by inducing regulation allowing automated
48
vehicles to travel only in motorways or by complicating and subsequently
49
delaying technological development in this field,
o unexpected incidents like fatal accidents; bankruptcy or change in strategic
1
priorities of major industry players could significantly influence the development
2
path for automated vehicles not only in the Netherlands, but also in any country,
3
o development of automated vehicles is expected to have implications on mobility
4
in all scenarios. These implications vary from very important in the ‘AV …in
5
bloom’ scenario to minimal in the ‘AV …in doubt’ scenario,
6
o the Dutch government is expected to take measures (e.g. travel demand
7
management) to curb growth of travel and subsequent externalities in three out of
8
the four scenarios.
9
In the last step of our study twenty experts assessed the likelihood of all
10
scenarios and overall impact (i.e., value of time, road capacity, and total VKT).
11
Scenario 2 (AV ….in bloom) and scenario 3 (AV …in demand) were perceived as the
12
most likely to happen in the future, while likelihood of scenario 1 (AV …in standby)
13
and scenario 4 (AV …in doubt) was assessed lower. The results show that these
14
experts expect policies in the Netherlands to be generally supportive of automated
15
vehicles (scenarios 2 and 3 exhibited the highest likelihood). Thus, it is probably
16
technology that will mainly influence development path of automated vehicles. In
17
fact, participants seem also to believe that technology has good chances to develop
18
rapidly and full automation to be a reality within the next ten years, if we take into
19
account that ‘AV …in bloom’ was ranked as the most likely scenario. It should be
20
noted that the average sum of probabilities per participant was 137.5%, while only
21
one person estimated a sum of probabilities below 100%. This indicates that the
22
participants were quite confident that the four scenarios can adequately describe
23
potential development paths of automated vehicles in the Netherlands or at least they
24
did not consider an alternative fifth scenario as more likely. Finally, the experts
25
expect the ‘AV …in bloom’ scenario to have the highest and the ‘AV …in doubt’
26
scenario the lowest overall impact.
27
In conclusion, our study suggests that fully automated vehicles will likely be a
28
reality between 2025 and 2045 and are expected to have significant implications for
29
mobility and planning policies in the Netherlands. The pace of development and
30
subsequent implications largely depend on technological evolution, policies and
31
customers’ attitude. These driving forces could be weighted differently in other
32
countries or from other experts with respect to their impacts and uncertainty. Thus,
33
the paths we identified for the development of automated vehicles in this study are
34
plausible but not the only ones. Moreover, our assessment exercise offers a rough
35
order of magnitude estimate of the possible impacts of automated vehicles in various
36
scenarios based on experts’ responses. Such estimations can offer input to subsequent
37
modeling exercises to explore these impacts and their complex interactions more
38
precisely. Sensitivity analysis in these modeling exercises could address uncertainty
39
about the magnitude of those effects.
40 41 42 ACKNOWLEDGEMENTS 43 44
This research was funded by the PBL Netherlands Environmental Assessment
45 Agency. 46 47 48 49 50
REFERENCES 1
2
1. Milakis, D., van Arem, B., van Wee B., 2015. The ripple effect of automated
3
driving. Paper presented at the BIVEC-GIBET Transport Research Day 2015.
4
Eindhoven, The Netherlands 28-29 May 2015.
5
2. National Highway Traffic Safety Administration, 2013. Preliminary statement of
6
policy concerning automated vehicles.
7
3. SAE International, 2014. Taxonomy and Definitions for Terms Related to
On-8
Road Motor Vehicle Automated Driving Systems. Warrendale, PA: SAE
9
International.
10
4. Underwood, S.E., 2014. Automated vehicles forecast: Vehicle Symposium
11
Opinion Survey. Ann Arbor, MI: Graham Institute for Sustainability, University
12
of Michigan.
13
5. Litman, T., 2014. Autonomous Vehicle Implementation Predictions Implications
14
for Transport Planning. Victoria Transport Policy Institute.
15
6. Kyriakidis, M., Happee, R., Winter, J.C.F. De, 2015. Public opinion on automated
16
driving: Results of an international questionnaire among 5,000 respondents.
17
Transportation Research Part F: Traffic Psychology and Behaviour 32: 127-140.
18
7. Hoogendoorn, R., Arem, B. Van, Hoogendoorn, S., 2014. Automated Driving,
19
Traffic Flow Efficiency And Human Factors: A Literature Review. Transportation
20
Research Record 2422: 113–120.
21
8. Van Arem, B., Smits, C., 1997. An exploration of the development of automated
22
vehicle guidance systems. Delft, The Netherlands: TNO.
23
9. VanderWerf, J., Shladover, S.E., Miller, M.A., Kourjanskaia, N., 2002. Evaluation
24
of the Effects of Adaptive Cruise Control Systems on Highway Traffic Flow
25
Capacity and Implications for Deployment of Future Automated Systems.
26
Transportation Research Record 1800: 78–84.
27
10. Arnaout, G., Bowling, S., 2011. Towards reducing traffic congestion using
28
cooperative adaptive cruise control on a freeway with a ramp. Journal of Industrial
29
Engineering and Management 4 (4): 699–717.
30
11. Calvert, S.C., Van Den Broek, T.H. A, Van Noort, M., 2011. Modelling
31
cooperative driving in congestion shockwaves on a freeway network. In: IEEE
32
Conference on Intelligent Transportation Systems, Proceedings, ITSC. 614–619.
33
12. Shladover, S.E., Su, D., Lu, X.-Y., 2012. Impacts of Cooperative Adaptive Cruise
34
Control on Freeway Traffic Flow. Transportation Research Record 2324: 63–70.
35
13. Malokin, A., Circella, G., Mokhtarian, P.L., 2015. How Do Activities Conducted
36
while Commuting Influence Mode Choice? Testing Public Transportation
37
Advantage and Autonomous Vehicle Scenarios. In: 94th Annual Meeting of the
38
Transportation Research Board.
39
14. Gucwa, M., 2014. Mobility and Energy Impacts of Automated Cars. Stanford,
40
CA: Department of Management Science and Engineering, Stanford University.
41
15. Fagnant, D.J., Kockelman, K.M., 2014. The travel and environmental implications
42
of shared autonomous vehicles, using agent-based model scenarios.
43
Transportation Research Part C: Emerging Technologies 40: 1–13.
44
16. International Transport Forum, 2015. Urban mobility: system upgrade. Paris:
45
OECD/International Transport Forum.
46
17. Silberg, G., Wallace, R., Matuszak, G., Plessers, J., Brower, C., Subramanian, D.,
47
2012. Self-driving cars: The next revolution. KPMG: Center for Automotive
48
Research.
18. Fagnant, D.J., Kockelman, K., 2015. Preparing a Nation for Autonomous
1
Vehicles: Opportunities, Barriers and Policy Recommendations for Capitalizing
2
on Self-Driven Vehicles. Transportation Research Part A: Policy and Practice 77:
3
1–20.
4
19. Wagner, J., Baker, T., Goodin, G., Maddox, J., 2014. Policy Implications of
5
Automated Vehicles on Texas Highways. College Station, Texas: Texas A&M
6
Transportation Institute
7
20. Anderson, J.M., Kalra, N., Stanley, K.D., Sorensen, P., Samaras, C., Oluwatola,
8
O.A., 2014. Autonomous Vehicle Technology. A Guide for Policymakers. Santa
9
Monica, CA: RAND.
10
21. Mahmoud, M., Liu, Y., Hartmann, H., Stewart, S., Wagener, T., Semmens, D.,
11
Stewart, R., Gupta, H., Dominguez, D., Dominguez, F., Hulse, D., Letcher, R.,
12
Rashleigh, B., Smith, C., Street, R., Ticehurst, J., Twery, M., van Delden, H.,
13
Waldick, R., White, D., Winter, L., 2009. A formal framework for scenario
14
development in support of environmental decision-making. Environmental
15
Modelling & Software 24 (7): 798–808.
16
22. Maack, J., 2001. Scenario Analysis: A Tool for Task Managers. In: Social
17
Development Paper no. 36. Social Analysis: Selected Tools and Techniques.
18
Washington DC: World Bank, 62–87.
19
23. Townsend, A., 2014. Re-programming mobility. The Digital Transformation of
20
Transportation in the United States. New York, NY: Rudin Center for
21
Transportation Policy & Management.
22
24. Schwartz, P., 1996. The art of the long view: Planning for the future in an
23
uncertain world. New York: Currency Doubleday.
24
25. Bood, R., Postma, T., 1997. Strategic learning with scenarios. European
25
Management Journal 15 (6): 633–647.
26 27
