From Pilot to Implementation: What are Potential Deployments with Automated Vehicles in Public Transport Based on Knowledge Gained from Practice?

Abstract

Automated vehicles (AVs) have the potential to complement our current public transport system and thereby improve the livability and accessibility. Many pilots have been conducted in the Netherlands, Europe, America and other countries worldwide, but permanently operating services are very limited. This paper aims to provide insight into possibilities and requirements of operational AV services. In order to develop these insights, we made an inventory of pilots from Europe in particular and conducted in-depth research of pilots in the Netherlands. Furthermore, some pilots of America are included.This insight can help public transport companies, (local/regional) government and public transport authorities to determine future, permanent, application of AVs. The main research question of this paper is: “What are the characteristics of promising situations where automated vehicles can be deployed in public transport based on knowledge gained from practice?” We conclude that the operating environment, the operational speed of the AVs and the presence of an on board steward are key factors for an operational AV service. The type of environment where the AVs are operating is often a semi-controlled, publicly accessible environment such as a campus, stadium, office park or entertainment area. The speed of the vehicles during the pilots was low, with most vehicles operating below 21 km/h. All pilots (except the Rivium ParkShuttle) operated with steward. The findings in this paper reveal that implementation of AVs is still very limited. With the abovementioned requirements, the possible implementations of the vehicles are often limited to short distances, such as first-/last-mile transport.

Full text

From Pilot to Implementation: What are Potential Deployments with Automated Vehicles
1
in Public Transport Based on Knowledge Gained from Practice?
2
3
Reanne Boersma (Corresponding author)
4
Researcher SURF-STAD project
5
Department of Transport & Planning & Research Centre Sustainable Port Cities
6
Delft University of Technology & Rotterdam University of Applied Science
7
Stevinweg 1, 2628CN, Delft & Heijplaatstraat 23, 3089JB, Rotterdam, the Netherlands
8
Email: reanne.boersma@crow.nl
9
ORCiD: 0000-0002-0063-1051
10
11
Irene Zubin
12
PhD candidate
13
Department of Transport & Planning
14
Delft University of Technology
15
Stevinweg 1, 2628CN, Delft, the Netherlands
16
Email: I.Zubin@tudelft.nl
17
18
Bart van Arem
19
Full Professor
20
Department of Transport & Planning
21
Delft University of Technology
22
Stevinweg 1, 2628CN, Delft, the Netherlands
23
Email: B.vanarem@tudelft.nl
24
ORCiD: 0000-0001-8316-7794
25
26
Niels van Oort
27
Assistant professor
28
Department of Transport & Planning
29
Delft University of Technology
30
Stevinweg 1, 2628CN, Delft, the Netherlands
31
Email: N.vanoort@tudelft.nl
32
ORCiD: 0000-0002-4519-2013
33
34
Arthur Scheltes
35
Consultant Public Transport & Automated Vehicles
36
Goudappel Coffeng
37
Anna van Beurenplein 46, 2595DA, the Hague, the Netherlands
38
Email: ascheltes@goudappel.nl
39
40
Frank Rieck
41
Lector Smart e-mobility
42
Research Centre Sustainable Port Cities
43
Rotterdam University of Applied Science
44
Heijplaatstraat 23, 3089JB, Rotterdam, the Netherlands
45
Email: f.g.rieck@hr.nl
46
47
48
Word Count: 5794 words + 4 figures (1000) + 2 tables (500) = 7294 words
49
50
Submitted July 2020
51

Boersma, Zubin, van Arem, van Oort, Scheltes, Rieck
2
ABSTRACT
1
Automated vehicles (AVs) have the potential to complement our current public transport system
2
and thereby improve the livability and accessibility. Many pilots have been conducted in the Netherlands,
3
Europe, United States and other countries worldwide, but permanently operating services are very limited.
4
This paper aims to provide insight into possibilities and requirements of operational AV services. In order
5
to develop these insights, we made an inventory of pilots from Europe in particular and conducted in-
6
depth research of pilots in the Netherlands. Furthermore, some pilots in the United States are
7
included.This insight can help public transport companies, (local/regional) government and public
8
transport authorities to determine future, permanent, application of AVs. The main research question of
9
this paper is: “What are the characteristics of promising situations where automated vehicles can be
10
deployed in public transport based on knowledge gained from practice?”
11
We conclude that the operating environment, the operational speed of the AVs and the presence
12
of an on board steward are key factors for an operational AV service. The type of environment where the
13
AVs are operating is often a semi-controlled, publicly accessible environment such as a campus, stadium,
14
office park or entertainment area. The speed of the vehicles during the pilots was low, with most vehicles
15
operating below 21 km/h. All pilots (except the Rivium ParkShuttle) operated with steward. The findings
16
in this paper reveal that implementation of AVs is still very limited. With the abovementioned
17
requirements, the possible implementations of the vehicles are often limited to short distances, such as
18
first-/last-mile transport.
19
20
Keywords: Automated Shuttle, Automated Public Transport, Automated Vehicle
21

Boersma, Zubin, van Arem, van Oort, Scheltes, Rieck
3
1. INTRODUCTION
1
Automated vehicles (AVs) are seen as an emerging technology with the potential to change the
2
way we travel. Several studies show the benefits of automated driving, such as improved safety [1],
3
improved traffic flows [2] and an increase in capacity of existing traffic infrastructure [3]. Furthermore,
4
AVs may expand access to transportation to those who cannot drive or those who face significant barriers
5
to driving, such as people with permanent or temporary disabilities and elderly people [4]. Because of
6
these advantages one would think lots of AVs will be operating on our roads. However, permanent
7
applications with AVs are (still) scarce.
8
Stark et al [5] suggest that, in Germany, the most favourable pathway for AV’s in public transport
9
(PT) is as a feeder system for high capacity transportation and as a first-/last-mile solution. They describe
10
the favourable usecase as a system-integrated use case which is oriented toward intermodality. The
11
favourable use case contains partially fixed routes and is partially an on-demand system. The starting
12
point or destination is at a transportation hub, near home or the workplace, respectively. Stark also
13
concludes that more research is needed, for example, about where in relation to spatial context an
14
operation with AVs may be possible [5].
15
Haque and Brakewood carried out a synthesis and comparison on automated shuttle pilot projects
16
in the United States. They identified nineteen shuttle pilot projects and selected six pilot projects for in-
17
depth research. The selected cases are intended to be representative of their general location type. The
18
shuttle projects had in common that they were all deployed in relatedly dense urban areas. For example
19
downtowns, entertainment areas, stadiums, office parks, University campus and transit stations. Right of
20
way was either a designated lane, in mixed traffic or an off-street trail, depending on deployment [6].
21
Furthermore, findings of the Federal Transit Administration in their Transit Automation Research Plan
22
indicate potential applications for shuttles being circulator bus service – fixed route or flexible service
23
between two or more points – and connections to fixed route transit stations. They conclude that shuttle
24
vehicles are limited to certain operating environments due to their low speed. Examples of potential
25
operations are parking lots, busways, campuses, downtown districts and retirement communities [7].
26
An online survey of 38 cities conducted by Bloomberg in 2017 shows that most cities are
27
interested in operating AVs on the last-mile to bridge the gaps at the edges of transit systems. They
28
consider last-mile solutions as “low-hanging fruit”. In addition they found more variety in the places
29
cities are choosing for pilots. Examples of pilot places are technology parks, campus areas, urban renewal
30
districts and former Olympic sites. These places have in common that they isolate AVs from the rest of
31
the city. Bloomberg concludes: “So while trials are increasingly happening in cities, they aren’t yet
32
tackling the full challenge of navigating complex urban environments” [8].
33
There seems to be a knowledge gap between what is theoretically possible with AVs and what is
34
happening in practice. Milakis et al [9] concluded in their paper that more empirical studies about first-
35
order implications of vehicle automation are a priority as the technology evolves. Our paper aims to
36
provide insight into possibilities and requirements of operational AV services. In order to develop these
37
insights, we made an inventory of pilots from Europe in particular and conducted in-depth research of
38
pilots in the Netherlands. Many pilots have been conducted, but very few pilots have developed into
39
permanently operating services. The main research question of this paper is: “What are the characteristics
40
of promising situations where automated vehicles can be deployed in public transport based on
41
knowledge gained from practice?” The outline of the paper is as follows. Section 2 describes the methods
42
used for this study. Section 3 contains the results of our research which will be discussed in section 4.
43
Section 5 presents the conclusions of this paper.
44
45
2. METHODS
46
In the past years, automated vehicles have been the center of many pilots and demonstrations. In
47
the context of public transport, these pilots engage automated mini-buses as first-/last-mile option for
48
main facilities. The automated mini-buses covered in this paper are considered to be level four according
49
to the SAE levels. The SAE levels provide insight into the automated capabilities of a vehicle and are
50

Boersma, Zubin, van Arem, van Oort, Scheltes, Rieck
4
often reffered to worldwide. The SAE levels distinguish six levels ranging from zero to five, with zero
1
being no-automation and five fully autonomous. The AVs in this paper are considered to be level four,
2
because they are capable of operating autonomously on the given trip within the geographical area. The
3
steward in the vehicle is capable to override the autonomous system and manually control the vehicle.
4
The AVs are not considered level five, because level five vehicles must be capable of operating
5
autonomously unconditionally, under all road conditions without expectation that a steward/user will be
6
able to respond to a request to intervene [10].
7
An inventory of European pilots involving automated mini-buses can be found in the report
8
“Automated buses in Europe: An inventory of Pilots”, where the authors gathered information on the
9
current state of the art and state of research [11]. In order to get insight in the state of the art of AVs, an
10
overview is included in the appendix of the Autobus publication. The overview contains the following
11
columns: country, project, location, date, vehicle, capacity, speed, route, length, infrastructure, research,
12
more information and comments.
13
The inventory of pilots in Europe and in-depth research on pilots in the Netherlands were
14
analyzed for this paper. Also, some pilots conducted in the United States are being considered [6] [12].
15
The inventory of European pilots, the Dutch pilots and some US pilots were analyzed to answer the
16
abovementioned research question: “What are the characteristics of promising situations where automated
17
vehicles can be deployed in public transport based on knowledge gained from practice?” To answer this
18
research question, three aspects are being considered:
19
- The type of environment where the shuttle is operating;
20
- The average speed of the shuttle;
21
- The presence of a steward on-board.
22
These three aspects are closely related to the main research question because the type of
23
environment the shuttles are operating in and the average speed of the shuttles during the pilots, show the
24
practical limitations of the system. Also, the presence of the on-board steward negates the expected
25
reduction of total costs of ownership (due to the absence of a driver) [13]. An overview of costs per hour
26
of Dutch public transport busses conducted in 2015, shows that 51% of the operating costs are direct
27
personnel costs [15]. An cost-based analysis conducted by Bosch et al. shows that operating costs can
28
lead to financial savings due to the absence of the driver and the possibility to operate the fleet more
29
flexible. Other costs, such as cleaning of the vehicle, may increase and should be considered [14]. The
30
expected reduction of totals costs of ownership can affect the decision-making of transportation operators.
31
Firstly, desk research was performed to gather information. Extensive internet research has
32
provided the necessary literature and some example cases from the United States. Information on pilot
33
projects in Europe is based on previous research of the authors in collaboration with the Autobus project
34
[11]. The in-depth research on Dutch pilots is based on previous research of the authors.
35
Case studies were conducted to gather in-depth information on several pilot projects in the
36
Netherlands. The Dutch cases were selected based on available in-depth information and the possibility to
37
conduct interviews and visit the project locations. The authors believe that the differences and similarities
38
of the Dutch cases are representative for pilots in the Netherlands. The Appelscha case study has been
39
chosen because the vehicle operated in mixed traffic with very little infrastructural changes to facilitate
40
the vehicle. The WEpod case study was chosen because this project was the first to operate an AV on
41
public roads. Also, the WEpod project invested heavily in the technique of the vehicle, but ultimately the
42
WEpod was not able to fulfill the intended service. The third selected case study is the case of the Rivium
43
ParkShuttle. The ParkShuttle is unique because this fully operational system has been operating since the
44
‘90s on a dedicated track without a steward. Lastly, findings from the AV vs PT inventory are included.
45
The AV vs PT inventory contains interviews with twelve Dutch public transport authorities to find out
46
how they think AVs will affect public transport. The interviews were conducted in 2017-2018 and an
47
overview of all pilots in the Netherlands was created. The AV vs PT research provides insight in the
48
question on where AVs are of added value in the PT system [14]. By combining practical knowledge and
49

Boersma, Zubin, van Arem, van Oort, Scheltes, Rieck
5
comparing a selection of cases, the authors aim to provide insight in potential future operations with
1
automated shuttles to contribute to the process of moving from pilots to more permanent operations.
2
3
4
3. RESULTS
5
3.1 Results pilots in Europe
6
With the ParkShuttle in Schiphol being one of the first automated mini-buses realized in 1997,
7
several transport companies started to get interested in this new technology, especially from 2016, the
8
year in which pilots started to gain momentum. By 2019, 118 pilots were performed in Europe, involving
9
18 countries and several research institutes. The companies Navya and EasyMile have provided almost
10
80% of these pilots with their autonomous shuttle Navya Arma and EasyMile EZ10, followed by Local
11
Motor’s Olli. The vehicles have a smaller size, smaller capacity and drive at a slower pace compared to
12
traditional buses, allowing for up to 15 passengers (including the steward) and driving at a maximum
13
operating speed of 25 km/h. The purpose of these pilots is to provide a last-mile connection to main
14
facilities such as university campuses, business and shopping areas, airports and parking facilities,
15
meaning that the route length is usually short, with most pilots not exceeding 1,5 kilometers per trip.
16
Despite the substantial amount of pilots, very limited research was found, with 33% of the pilots
17
having a detailed documentation and only 8.5% of pilots supported by proper research. Consequently, one
18
of the most insightful sources for potential deployments of automated mini-buses can be found in the
19
practical demonstrations of these pilots, focusing on the knowledge gained from practice. Among the 18
20
European countries where pilots took place, the Netherlands is one of the leading countries, with 8 pilots
21
conducted from 1997 to 2019 [11]. For this reason, the following paragraphs focus on three case studies
22
from the Netherlands: Appelscha, WEpod in the province of Gelderland and Rivium ParkShuttle in
23
Rotterdam.
24
25
3.2 AVs on a bicycle lane in Appelscha
26
The municipality of Ooststellingwerf ran a pilot with automated vehicles on the bicycle lane near
27
Appelscha (NL) in the fall of 2016. The predicted decline of inhabitants and the already shrinking public
28
transport network prompted the project. The municipality rented two electric, automated Easymile EZ10
29
vehicles to preserve the region’s accessibility. These vehicles operated with a steward on the separate
30
cycle lane just outside the city limits. The maximum speed of the vehicles was 15 km/h. To prepare the
31
separate cycle lane, relatively small adjustments were made. Such as overtaking bays, warning signs,
32
temporary matrix signs and a temporary, yellow yield line was placed to give way to the users on the
33
cycle lane. During the project the vehicle detected all sorts of obstacles next to the cycle lane such as
34
garden fences, high grass and low tree branches. Therefor, the vehicle had to move 20 centimeters (the so-
35
called ‘virtual space’) away from the edge of the cycle lane. The width of the cycle lane varied between
36
2.70 m and 3.10 m. The width of the vehicle was 1.99 m. The remaining space for the cyclists to pass or
37
overtake the vehicle varied between 0.71 m and 1.11 m. CROW guidelines (commonly used in road
38
design and well accepted in the Netherlands) advises a minimal space of 1.20 m for cyclists to
39
comfortably cycle past objects or, in this case, the AV.
40
The main finding of this project is that a pilot with AVs is possible with little infrastructural
41
changes, but the project also shows that this particular cycle lane was not sufficient due to the width of the
42
cycle lane [15]. Figure 1 shows the vehicle in operation on the cycle lane near Appelscha.
43

Boersma, Zubin, van Arem, van Oort, Scheltes, Rieck
6
1
Figure 1: Appelscha vehicle in operation [15]
2
3
3.3 AV on Public Road in Wageningen
4
The WEpod project was initiated by the province of Gelderland in 2014 and was the first project
5
with AVs on public roads in the Netherlands. The province of Gelderland has the ambition to provide
6
new, flexible, sustainable and social mobility. The WEpod project was initiated to give an impulse to
7
regional economy and education, as well as gaining knowledge of operating an automated vehicle. This
8
knowledge can spark discussions regarding smart public transport solutions for public transport
9
companies, government, road authorities, businesses and society in general. Two Easymile EZ10 vehicles
10
were purchased and adapted for the project. In addition to the regular hard- and software of the vehicles,
11
the project team added radars, camera’s, LIDAR’s, computers, intercoms and a seat for the steward (this
12
enumeration is not exhaustive). The radars and cameras form the “high level safety system” and is able to
13
detect and classify all objects around the vehicle. In addition to the high level safety system, the WEpod is
14
equipped with the low level safety system. This system serves to support the high level safety system and
15
is very reliable. However, the low level safety system is less intelligent than the high level safety system.
16
When the high level safety system misses an object, the low level system will intervene. Since the low
17
level safety system is less intelligent, this system has more “false positives” (meaning it sometimes
18
detects objects that are not there). The low level safety system consists of lasers (LIDAR) and ultrasonic
19
sensors and can be seen as a kind of virtual bumper with a detection zone up to 50 meters, depending on
20
the speed of the vehicle. Safety zones are set in advance. These safety zones have a certain size dependent
21
on the required braking distance in relation to the speed, taking additional safety margins into account.
22
The vehicle will initiate an emergency stop as soon as an object in the safety zone is detected. The picture
23
below shows the safety zones of the WEpods [16].
24

Boersma, Zubin, van Arem, van Oort, Scheltes, Rieck
7
1
Figure 2: WEpod safety zones [16]
2
The green zone on the top right indicates the range of the lasers. The bar at the bottom left
3
indicates when an emergency stop is initiated. An object in the green zone will not initiate an emergency
4
break. An object in the orange zone will slow the vehicle down and an object in the red zone will initiate
5
an emergency break. A steward was always present to monitor the vehicle and to activate the emergency
6
break if needed.
7
The WEpod project contained two routes; a route on the campus of the University of Wageningen
8
in the foodvalley and a route from the train station Ede/Wageningen to the campus. To prepare the route
9
some infrastructural changes deemed necessary. Some of these changes were imposing a parking ban and
10
speed limits. The already existing traffic light on the route was equipped with wifi-p to connect to the
11
vehicles. An additional traffic light was installed at the crossing of the Emmalaan near the train station.
12
This traffic light was only activated when the vehicle approached. Furthermore, a bus stop was created at
13
the Ede/Wageningen train station. The maximum speed of the WEpod was 25 km/h.
14
Some of the main findings of this project are the challenges in the exemption procedure and the
15
technical challenges of operating the vehicle. Since this was the first project with AVs on public roads, an
16
exemption from RDW was mandatory. By going through the procedure first, a legal path was created for
17
other pilots. Also, the technical challenges with regards to operating the vehicle deemed greater than
18
expected. Even though the project team added hard- and software to upgrade the vehicle and make it more
19
intelligent, the service they had envisioned from the Ede/Wageningen train station to the campus has not
20
been realized [16].
21
22
3.4 AV shuttle without steward on dedicated road: Rivium ParkShuttle
23
The Rivium ParkShuttle operates without a steward on board, on a designated lane at the Rivium
24
Businesspark near Rotterdam (NL). To fulfil the last mile from the metro station Kralingse Zoom to
25
businesspark Rivium, the municipality of Capelle aan den IJssel and other stakeholders invested in
26
automated electric people movers from 2getthere. The experimental phase started in 1997 and from 1999
27
until 2001 the first generation ParkShuttles were in operation. The second generation ParkShuttles have
28

Boersma, Zubin, van Arem, van Oort, Scheltes, Rieck
8
been operating since 2005. The Rivium ParkShuttle is unique since it is (to the knowledge of the authors)
1
the only operational automated vehicle in Europe in permanent (revenue generating) service. Having been
2
operational for over 10 years, the ParkShuttle operation may be considered ‘proven technology’.
3
The ParkShuttle operates on a designated lane consisting of two-lanes, except the overpass over
4
the Abraham van Rijckevorselweg and the tunnel near the metro station. The lane was initially designed
5
as a bicycle lane, hence the one-way viaduct and tunnel. Crossing traffic is managed with barriers (see
6
figure 3). Although there is an occasional cyclist or pedestrian using the dedicated infrastructure of the
7
ParkShuttle, there is minimal interaction with other traffic. The Frog system is integrated in the
8
infrastructure which means the asphalt contains magnetic landmarks to help the vehicles navigate. The
9
operating speed of the vehicle is 32 km/h [17].
10
11
Figure 3: Second generation Rivium ParkShuttle [17]
12
The main finding of this case study is that the Rivium ParkShuttle can be an example of
13
deployment of AVs in the PT system. The main benefit of this dedicated infrastructure is the lack of
14
interruptions so that the timetable can be optimally executed. The third generation ParkShuttles are
15
getting ready and will be ‘smarter’ than the second generation. The new ParkShuttle will operate partly on
16
public roads (±300 m), making this the first automated vehicle on public roads without a steward on
17
board. An operator will be able to intervene from the control room. This extension of the route and
18
deployment of the third generation vehicle is planned for 2020/2021 [17].
19
20
3.5 Expert opinion of public transport authorities in regards to AVs in PT
21
To gain insight into possibilities with AVs in the current PT system, interviews were conducted
22
with twelve Dutch public transport authorities [14]. During the interviews, questions were asked about
23
threats/opportunities as well as feasibility, visions and knowledge gaps with regards to operating AVs in
24
the current PT system. Subsequently questions were asked about what the future of PT would look like.
25
The interviews were conducted in 2017 – 2018.
26
In general, the interviewed PT authorities agree that PT on main lines will remain the same. The
27
PT authorities expect more possibilities with AVs as first-/last-mile solution to and from main lines. The
28
main challenges for PT is the accessibility, the aging society and declining population (mostly in rural
29
areas) as well as keeping PT feasible. An example that shows the possibilities of AVs on the first- and
30
last-mile near Delft-Zuid train station, is a study conducted by Scheltes et al. The simulations of Scheltes
31
et al. show different operational scenarios with a reduced average total travel time of over six minutes
32
[18]. Besides the possibilities of operating AVs as first-/last-mile solution, the PT authorities mentioned
33
the risk of drawing people from active modes, such as walking and cycling, to the automated transport
34
system. A study conducted by Hezaveh et al. shows the possible impact on the network due to excessive
35
demand and the risk of people moving from active modes to automated vehicles. Also, the study shows
36
that automated PT might remove mobility barriers for captive drivers, which will increase demand [19].
37

Boersma, Zubin, van Arem, van Oort, Scheltes, Rieck
9
In general the PT authorities agreed that the first-/last-mile is the most beneficial for AVs at this
1
time. They believe first-/last-mile routes are most beneficial because of the current low speed of the
2
vehicles (max 32 km/h). The PT authorities also noted that current projects with AVs are mostly focused
3
on the technical aspects. However, the challenges regarding the deployment of AVs extend beyond the
4
technical level. The interviewed parties indicate that it is important to focus, with the upcoming pilots,
5
more on the traveler and the position of the vehicle within the existing PT network. The interviewed
6
parties stress that it is important to think about the long-term implementation [14].
7
8
4. Analysis of operational requirements
9
The findings reported in this paper provide insight in potential situations where AVs can be
10
deployed in public transport. Three aspects are being considered in this paper:
11
• The type of environment where the shuttle is operating;
12
• The average speed of the shuttle;
13
• The presence of a steward on-board.
14
15
4.1 Type of environment
16
When looking at the type of environment of previous projects, limitations of the AVs can be
17
determined. The infrastructural adaptations derived from the European pilots mainly concern road
18
markings, warning signs, installation of equipment for V2X communication (sensors, traffic lights etc.)
19
and temporary bus stops. Some similar infrastructural changes were found in the Dutch pilots. The Dutch
20
pilots showed the following infrastructural changes:
21
22
Appelscha
WEpod
Rivium ParkShuttle
- Overtaking bays
- Warning signs
- Temporary matrix
signs
- Temporary yellow
yield line
- Parking ban
- Speed limit
- Existing traffic light
equipped with wifi-p
- Traffic light installed
- Bus stop
- Dedicated lane
(equipped with
magnets)
- Barriers for crossing
traffic
- Control room
- Bus stops
The Appelscha and WEpod pilots were conducted with relatively little infrastructural changes,
23
similar to the findings of the European pilots. For the Rivium ParkShuttle a dedicated lane was
24
constructed (at first designed to be a bicycle lane). This means major infrastructural adjustments, but very
25
little interruptions in operation. Also, in contrast to the Appelsha and WEpod pilots, the Rivium
26
ParkShuttle is, as far as the authors know, the only permanently operating shuttle system in Europe.
27
Other than the infrastructural adjustments, the environment of the vehicles provide insight into
28
promising situations for deployment of AVs. Most European pilots served as a first-/last-mile solution
29
between PT stops/stations and campus areas, business or shopping districts or within airports, parking
30
facilities or city centres [11]. These environments are often limited accessible for other road users Also,
31
during the interviews conducted with PT authorities it was clear that they believe AVs are likely to fulfil
32
the first-/last-mile in the current PT network. They also indicated that pilots in the Netherlands are often
33
focused on technical capabilities and are restricted due to the low speed of the vehicle [14].
34
35
4.2 Speed of the shuttle
36
78% of 82 pilots in Europe have an average speed below 21 km/h. Two pilots exceeded 40 km/h.
37
Most pilots were limited in the allowed speed and the average operational speed was often lower than the
38
design speed [11]. Figure 4 shows the average operational speed of the pilots in Europe.
39

Boersma, Zubin, van Arem, van Oort, Scheltes, Rieck
10
1
Figure 4: Number of pilots per average operational speed of the AVs in km/h [11]
2
The Appelscha, WEpod and Rivium ParkShuttle operate with 15 km/h, 25 km/h and 32 km/h.
3
The Appelscha project has shown that if the vehicle is operating too slow on the bicycle lane, bicycles
4
will try to overtake the vehicle [15]. This could lead to potentially dangerous situations. The WEpod
5
project has provided more insight in the high- and low-level safety systems and shows that a cyclist
6
overtaking the vehicle can be seen as an object which can trigger the emergency break [16]. This sudden
7
stop can be confusing for other cyclists or road users.
8
The synthesis and comparison of pilots in the United States indicates that the average speed of the
9
shuttles was between 16-24 km/h [6]. Also, the Automated Mobility District Implementation Catalog
10
provides insight in nine shuttle projects in the United States. The nine selected projects have in common
11
that they all operate with low speed varying from 8-23 mph, which is approximately 12-37 km/h [12].
12
13
4.3 Presence of on-board steward
14
The European pilot overview and the Dutch cases found that most pilots were conducted with on-
15
board steward, except the Rivium ParkShuttle [11] [15] [16] [17]. Also, the previously mentioned studies
16
in regards to pilot projects in the United States all show operation with a steward inside the vehicle [6]
17
[12]. This probably has to do with legislation. The Convention on Road Traffic, also known as Vienna
18
Treaty, in force in Europe, stipulates that “every moving vehicle or combination of vehicles shall have a
19
driver” [20]. Dutch law describes the driver as the person who drives the vehicle or who is deemed to
20
have the vehicle driven under his immediate supervision [21] and stipulates that the driver must be able to
21
bring the vehicle to a complete stop [22]. The new experimental law, implemented on the 1st of July 2019,
22
allows tests with a remote driver on public roads. At the time of writing this paper, there are no pilots
23
planned based on an exemption under the Experimental law [23].
24
25
5. DISCUSSION AND CONCLUSION
26
AVs have the potential to complement our current PT system and thereby improve the livability
27
and accessibility. Many pilots have been conducted in the Netherlands, Europe, United States and other
28
countries worldwide. It is remarkable that very few places are permanently operating AVs. One of the
29
reasons for doing this research was to find out why permanent operations with AVs is scarce and to find
30
out what is needed to progress the development from pilots to operations. This paper aims to provide
31
insight in practical applications with AVs and to fulfill the knowledge gap with regards to knowledge
32
gained from practice. This insight can help public transport companies, municipality, local/regional
33
government and public transport authorities etc. to determine future, perhaps permanent, application of
34
AVs.
35

Boersma, Zubin, van Arem, van Oort, Scheltes, Rieck
11
The main research question of this paper is: “What are the characteristics of promising situations
1
where automated vehicles can be deployed in public transport based on knowledge gained from practice?”
2
To answer this question information was gathered from literature and previous studies conducted by the
3
authors. Three aspects are being considered:
4
- The type of environment where the shuttle is operating;
5
- The average speed of the shuttle;
6
- The presence of a steward on-board.
7
The findings in this paper reveal that implementation of AVs as operational system is still very
8
limited. The type of environment where the AVs are operating, is often a semi-controlled, publicly
9
accessible environment such as a campus, stadium, office park or entertainment area. Some AVs operated
10
in mixed traffic situations with some infrastructural changes such as a changed priority situation and
11
temporary lines on the road. Three different scenarios based on the findings in this paper, are illustrated in
12
figure 5. The first scenario shows the AVs operating on a dedicated lane. The second scenario shows the
13
AV operating on public roads, but with the aforementioned conditions such as several warning signs and
14
limited interaction with other traffic. This situation is comparable with the semi-controlled, publicly
15
accessible environments such as campus areas or businessparks. The third scenario shows the AV
16
operating in mixed traffic on public roads, but with speed limits and a parking ban. The figure shows the
17
increase of complexity in which the AVs discussed in this paper are operating.
18
19
20
The speed of the vehicles during the pilots were low, respectively 8 km/h until a maximum of 40
21
km/h, with most vehicles operating below 21 km/h. Also, all pilots (except the Rivium ParkShuttle)
22
operated with a steward on-board. With these preconditions, the possible implementations of the vehicles
23
are often limited to short distances in (semi-)controllable, but publicly accessible environments. This
24
corresponds to the expectation that AVs are most beneficial as first- and last-mile solution, as mentioned
25
by interviewed transport authorities in our AV vs PT study and as mentioned by Bloomberg as “low
26
hanging fruit”.
27
28
Figure 5: Three situations of possible AV operations

Boersma, Zubin, van Arem, van Oort, Scheltes, Rieck
12
The pilots in the Netherlands, Europe and United States show a broad interest in implementing
1
AVs in PT worldwide. Technical issues are being fixed with every new pilot and the technology is
2
improving fast. In the Netherlands, there seems to be a shift from focussing on technical aspects to
3
focusing on fulfilling a PT gap and offering a service with AVs. Therefor, the goal of the pilots seems to
4
be shifting from short-term experiments to long-term pilots or even permanent applications. The
5
experimental law in the Netherlands creates the opportunity to experiment with AVs without on-board
6
steward. This could potentially stimulate the transition to operating AVs without steward on-board, which
7
might make operating AVs in PT more attractive for PT companies/authorities due to the absence of a
8
driver and thereby an expected reduction in total cost of ownership.
9
10
ACKNOWLEDGMENTS
11
The research in this paper was funded by the Spatial and Transport impacts of Automated Driving
12
(STAD) project (NWO project nr. 438-15-161), the Autobus project and Goudappel Coffeng.
13
14
AUTHOR CONTRIBUTIONS
15
The authors confirm contribution to the paper as follows: Study conception and design: RB, BvA, FR;
16
Data collection: RB, IZ; Analysis and interpretation of results: RB, IZ, BvA, NvO, FR; Draft manuscript
17
preparation: RB, IZ, AS. All authors reviewed the results and approved the final version of the
18
manuscript.
19
20
REFERENCES
21
22
[1]
D. Fagnant and K. Kockelman, "Preparing a nation for autonomous vehicles:
opportunities, barriers and policy recommendations," Transportation Research Part A,
2015.
[2]
R. E. Stern, S. Cui, M. L. D. Monache, R. Bhadani, M. Bunting, M. Churchill, N.
Hamilton, R. Haulcy, H. Pohlmann, F. Wu, B. Piccoli, B. Seibold, J. Sprinkle and D. B.
Work, "Dissipation of stop-and-go waves via control of autonomous vehicles: Field
experiments," Transportation Research Part C: Emerging Technologies, 2017.
[3]
B. Friedrich, "The Effect of Autonomous Vehicles on Traffic," in Autonomous
Driving, Springer, Berlin, Heidelberg, 2016, pp. 317-334.
[4]
S. H. Bayless and S. Davidson, "Driverless Cars and Accessibility - Designing the
Future of Transportation for People with Disabilities," The Itelligent Transportation Society
of America (ITS America), 2019.
[5]
K. Stark, K. Gade and D. Heinrichs, “What Does the Future of Automated Driving
Mean for Public Transportation?,” Transportation Research Record: Journal of the
Transportation Research Board, vol. 2673, no. 2, pp. 85-93, 2019.
[6]
A. M. Haque and C. Brakewood, "A synthesis and comparison of American
automated shuttle pilot projects," Case Studies on Transport Policy, 2020.
[7]
V. Valdes, G.-W. Torng, S. Mortensen and D. Diggs, "Stategic Transit Automation
Research Plan (0116)," 2018. [Online]. Available: https://www.transit.dot.gov/research-
innovation/strategic-transit-automation-research-plan-report-0116. [Accessed 25 June
2020].

Boersma, Zubin, van Arem, van Oort, Scheltes, Rieck
13
[8]
Bloomberg Philanthropies, "We spent 2017 scouring the globe to understand how
cities are preparing for AVs," The Aspen Institute, 2019. [Online]. Available:
https://avsincities.bloomberg.org/global-atlas/about/2017. [Accessed 29 June 2020].
[9]
D. Milakis, B. van Arem and B. van Wee, "Policy and society related implications
of automated driving: A review of literature and directions for future research," Journal of
Intelligent Transportation Systems, vol. 21, pp. 324-348, 2017.
[10]
SAE International, Surface vehicle recommended practice - Taxonomy and
Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles,
2018.
[11]
M. Hagenzieker, A. Boersma, J. Nunez Velasco, M. Ozturker, I. Zubin and D.
Heikoop, "Automated buses in Europe: An inventory of Pilots, version: 1.0,"
Autobus/STAD, Delft, 2020.
[12]
S. Young and L. J. Sam, "The Automated Mobility District Implementation
Catalog: Insights from Ten Early-Stage Deployments," 2020. [Online]. Available:
https://www.nrel.gov/docs/fy20osti/76551.pdf. [Accessed 26 June 2020].
[13]
A. Ongel, E. Loewer, F. Roemer, G. Sethuraman, F. Chang and M. Lienkamp,
“Economic Assessment of Autonomous Electric Mircrotransit Vehicles,” Sustainability,
vol. 11, no. 648, 2019.
[14]
P. M. Bösch, F. Becker, H. Becker and K. W. Axhausen, "Cost-based analysis of
autonomous mobility services," Transport Policy, vol. 64, pp. 76 - 91, 2018.
[15]
R. Boersma, A. Scheltes and N. van Oort, "Automatische voertuigen; kans of een
bedreiging voor het OV in Nederland?," in Colloquium Vervoersplanologisch Speurwerk,
Amersfoort (NL), 2018.
[16]
R. Boersma, B. van Arem and F. Rieck, "Application of Driverless Electric
Automated Shuttles for Public Transport in Villages: The Case of Appelscha," World
Electric Vehicle Journal, vol. 9, p. 15, 2018.
[17]
R. Boersma, B. van Arem and F. Rieck, "Casestudy WEpod: een onderzoek naar de
inzet van automatisch vervoer in Ede/Wageningen," 2018.
[18]
R. Boersma, D. Mica, B. van Arem and F. Rieck, "Driverless electric vehicles at
Businesspark Rivium near Rotterdam (the Netherlands): from operation on dedicated tracks
since 2005 to public roads in 2020," EVS31, Kobe, 2018.
[19]
A. Scheltes and G. Homem de Almeida Correia, "Exploring the use of automated
vehicles as last mile connection of train trips through an agent-based simulation model: An
application to Delft, Netherlands," International Journal of Transportation Science and
Technology, vol. 6, no. 1, pp. 28-41, 2017.
[20]
A. M. Hezaveh, C. Brakewood and C. R. Cherry, "Exploring the effect of
autonomous vehicles on transit ridership," in Transportation Research Record,
Washington, DC, USA, 2019.
[21]
Article 8 Convention on Road Traffic, Vienna, 1968.

Boersma, Zubin, van Arem, van Oort, Scheltes, Rieck
14
[22]
Article 1 paragraph 1 sub a Wegenverkeerswet 1994 (in Dutch).
[23]
Article 19 Reglement verkeersregels en verkeerstekens 1990 (in Dutch).
[24]
Government of the Netherlands, "Green light for Experimental Law for testing self-
driving vehicles on public roads," 2 July 2019. [Online]. Available:
https://www.government.nl/latest/news/2019/07/02/green-light-for-experimental-law-for-
testing-self-driving-vehicles-on-public-
roads#:~:text=On%201%20July%2C%20the%20new,types%20of%20experiments%20is%
20enacted.&text=The%20Netherlands%20has%20allowed%20publ. [Accessed 29 June
2020].
1
2