Content uploaded by Dimitris Milakis
Author content
All content in this area was uploaded by Dimitris Milakis on Dec 31, 2015
Content may be subject to copyright.
Milakis, Snelder, van Arem, van Wee, Correia
!
1
Scenarios about development and implications of automated 1!
vehicles in the Netherlands 2!
3!
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!
Milakis, Snelder, van Arem, van Wee, Correia
!
2
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!
Milakis, Snelder, van Arem, van Wee, Correia
!
3
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 50!
Milakis, Snelder, van Arem, van Wee, Correia
!
4
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 50!
Milakis, Snelder, van Arem, van Wee, Correia
!
5
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!
Milakis, Snelder, van Arem, van Wee, Correia
!
6
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!
Milakis, Snelder, van Arem, van Wee, Correia
!
7
!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 52!
Milakis, Snelder, van Arem, van Wee, Correia
!
8
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 50!
Milakis, Snelder, van Arem, van Wee, Correia
!
9
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!
Milakis, Snelder, van Arem, van Wee, Correia
!
10
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 50!
Milakis, Snelder, van Arem, van Wee, Correia
!
11
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. 49!
Milakis, Snelder, van Arem, van Wee, Correia
!
12
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!
!33!
Milakis, Snelder, van Arem, van Wee, Correia
!
13
!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!
Milakis, Snelder, van Arem, van Wee, Correia
!
14
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, 50!
Milakis, Snelder, van Arem, van Wee, Correia
!
15
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!
Milakis, Snelder, van Arem, van Wee, Correia
!
16
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. 49!
Milakis, Snelder, van Arem, van Wee, Correia
!
17
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!