Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges

Abstract

Automation is expected to effectively address the growing demand for passenger and freight transportation, safety issues, human errors, and increasing congestion. The growth of autonomous vehicles using the state-of-the-art connected vehicle technologies has paved the way for the development of passenger and freight autonomous trains (ATs), also known as driverless trains. ATs are fully automated trains that are centrally controlled using advanced communication and internet technologies, such as high-speed internet (5G) technology, Internet of Things, dedicated short range communications, digital video detection cameras, and artificial intelligence-based methods. The current study focuses on a detailed up-to-date review of the existing trends, technologies, advancements, and challenges in the deployment of ATs with a full automation level in rail transportation. The basic AT features along with the key technologies that are instrumental for the AT deployment and operations are discussed in detail. Furthermore, a comprehensive evaluation of the state-of-the-art research efforts is performed as well with a specific emphasis on the issues associated with the AT deployment, user perception and outlook for ATs, innovative concepts and models that could be used for the AT design, and the AT operations at highway-rail grade crossings. Based on the conducted review, this study determines the main advantages and challenges from the AT deployment. The identified challenges have to be collaboratively addressed by the relevant stakeholders, including railroad companies, researchers, and government representatives, to facilitate the AT development and deployment considering the perspectives of future users and without affecting the safety level.

Full text

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
1
Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.Doi Number
Deployment of Autonomous Trains in Rail
Transportation: Current Trends and Existing
Challenges
Prashant Singh1, Maxim A. Dulebenets1, (Member, IEEE), Junayed Pasha1, Ernesto D.R.
Santibanez Gonzalez2, Yui-yip Lau3, and Raphael Kampmann1
1Department of Civil & Environmental Engineering, Florida A&M University-Florida State University (FAMU-FSU) College of Engineering, Tallahassee, FL 32310 USA
2Department of Industrial Engineering, University of Talca, Curicó, Maule, Chile
3Division of Business and Hospitality Management, The Hong Kong Polytechnic University, Hong Kong
Corresponding author: Maxim A. Dulebenets (e-mail: mdulebenets@eng.famu.fsu.edu).
This work was supported in part by the Florida Department of Transportation under grants BDV30-977-26 and BDV30-977-33. The opinions, findings and
conclusions expressed in this publication are those of the authors and not necessarily those of the Florida Department of Transportation or the U.S.
Department of Transportation.
ABSTRACT Automation is expected to effectively address the growing demand for passenger and freight
transportation, safety issues, human errors, and increasing congestion. The growth of autonomous vehicles
using the state-of-the-art connected vehicle technologies has paved the way for the development of
passenger and freight autonomous trains (ATs), also known as driverless trains. ATs are fully automated
trains that are centrally controlled using advanced communication and internet technologies, such as high-
speed internet (5G) technology, Internet of Things, dedicated short range communications, digital video
detection cameras, and artificial intelligence-based methods. The current study focuses on a detailed up-to-
date review of the existing trends, technologies, advancements, and challenges in the deployment of ATs
with a full automation level in rail transportation. The basic AT features along with the key technologies
that are instrumental for the AT deployment and operations are discussed in detail. Furthermore, a
comprehensive evaluation of the state-of-the-art research efforts is performed as well with a specific
emphasis on the issues associated with the AT deployment, user perception and outlook for ATs, innovative
concepts and models that could be used for the AT design, and the AT operations at highway-rail grade
crossings. Based on the conducted review, this study determines the main advantages and challenges from
the AT deployment. The identified challenges have to be collaboratively addressed by the relevant
stakeholders, including railroad companies, researchers, and government representatives, to facilitate the
AT development and deployment considering the perspectives of future users and without affecting the
safety level.
INDEX TERMS Autonomous trains, driverless trains, connected vehicles, autonomous vehicles,
automation trends, automation challenges.
I. INTRODUCTION
Despite the marginal impacts of COVID-19 on rail freight
and passenger volume globally, the share of rail traffic is
expected to grow [1]. The implementation of new
technologies and better infrastructure will provide much
needed impetus to the rail transportation sector. The
statistics shows that the rail network has seen a tremendous
growth globally in the last two decades. The total length of
rail miles increased from 1,099,685 km in 2004 to
1,142,890 km in 2018 (see Fig. 1), which is approximately
a 4% increase [2]. The growth has been observed in terms
of the number of passengers traveling and in terms of
freight movements as well (see Fig. 2). More specifically,
the rail passenger traffic around the world increased from
2,440,732 million passenger-km in 2004 to 4,068,548
million passenger-km in 2018. On the other hand, the rail
freight traffic around the world increased from 8,443,020
million ton-km in 2004 to 11,190,112 million ton-km in
2018.
TABLE I shows the distribution of rail passenger and
freight traffic as well as the distribution of rail lines

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
2
globally by major geographical regions [2]. The data show
that Asia and Oceania (excluding Russia and Turkey) has
the largest rail passenger traffic accounting for 76.1% of the
total rail passenger traffic. Europe (including Turkey),
Russia, and Africa account for 15.7%, 4.4%, and 2.0% of
the total rail passenger traffic, respectively. The lowest rail
passenger traffic is observed in America (only 1.8% of the
total rail passenger traffic). As for the rail freight traffic,
Asia and Oceania, America, and Russia accumulate the
largest amount of rail freight traffic accounting for 38.0%,
31.8%, and 21.5% of the total rail freight traffic,
respectively. Furthermore, America has the largest rail
network (accounting for 34.1% of the total rail miles),
followed by Asia and Oceania, Europe, Russia, and Africa.
FIGURE 1. Changes in the total length of rail miles between 2004 and
2018.
The rail freight transportation market is expected to
achieve a cumulative average growth rate of around 2%
globally during the projected period of 2020 to 2025 [1].
Despite the fact that North America is recognized as a
global leader in rail freight market, it is anticipated that
Asia-Pacific is likely to overtake North America in the
following years. Although the growth of containerized
cargo has been boosted by the development of intermodal
transportation [3-5], non-containerized and liquid-bulk
freight still governs rail freight transport [1]. Furthermore,
the development of various cross-country and inter-
continental rail networks, such as Chongqing-Xinjiang-
Europe international railway, and other initiatives, such as
―One Belt One Road‖ by China, is expected not only to
provide a projected growth for rail freight and passenger
traffic but to open new economic corridors as well. In
particular, the ―One Belt One Road‖ initiative opens a
variety of economic corridors, including the following [6]:
(1) China-Mongolia-Russia economic corridor; (2) China-
Pakistan economic corridor; (3) Bangladesh-China-India-
Myanmar economic corridor; (4) China-Indochina
Peninsula economic corridor; (5) China-Central Asia-West
Asia economic corridor; and (6) New Eurasia Land Bridge
economic corridor.
FIGURE 2. Changes in the rail freight and passenger traffic between
2004 and 2018.
The growth in rail passenger and freight traffic along
with a continuous rail network expansion requires railroad
companies making improvements in the existing operations
to maintain profitability and effectively tackle the demand
for rail transportation. Similar to rail transportation, road
transportation is facing challenges associated with the
growing demand. The introduction of the connected and
autonomous vehicle (CAV) technology for road
transportation is expected to effectively address the
growing demand, safety issues, additional costs,
environmental problems, human errors, and increasing
roadway congestion. Some of the major benefits of having
the autonomous vehicle (AV) technology in transportation
include better accessibility, mobility, and improvement in
land use. A simultaneous use of the AV technology with
electric vehicles can significantly reduce the emissions and
protect the environment. Moreover, the deployment of AVs
is expected to reduce the vehicle ownership pattern. Fig. 3
shows the projected growth of driverless vehicles and
illustrates the following three eras [7]: (a) ―Era 1‖ – AVs
are being developed for consumers; (b) ―Era 2‖ –
consumers begin adopting AVs; and (c) ―Era 3‖ – AVs
become the primary means of transport. The AVs are
generally classified into six levels of automation [8]. Levels
―0‖, ―1‖, and ―2‖ require inputs from a driver, while levels
―3‖, ―4‖, and ―5‖ are mostly based on the auto pilot (the
driver input is required only by level ―3‖ upon request – see
Fig. 4).
Many countries within the European Union (EU) have
made a substantial progress to support the development of
cooperative, connected, and fully automated mobility. The

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
3
EU established a framework for Cooperative-Intelligent
Transportation Systems (C-ITS) that has the following main
objectives [9]: (1) develop a common vision for the EU
countries to combine the CAV development and
deployment efforts of different stakeholders; (2) launch a
C-Roads platform to coordinate the CAV deployment
across the EU; (3) address the legal issues associated with
the CAV deployment; (4) address the cybersecurity issues
associated with the CAV deployment; and (5) address the
data privacy issues associated with the CAV deployment.
The C-Roads program has been successfully implemented
in the Netherlands, Austria, Belgium, France, Germany,
United Kingdom, Nordic countries (i.e., Denmark, Finland,
Norway, and Sweden), and other EU countries as well. The
C-ITS framework supports the development of CAVs with
advanced features and technologies, such as hazardous
location notification, emergency vehicle approaching
notification, weather condition services, collision risk
warning, in-vehicle speed limits, green light optimal speed
advisory, and vulnerable road user protection [9]. Many of
the aforementioned CAV development tendencies and
technologies can be observed not only in the EU but also in
other countries as well (e.g., the United States, Australia,
Middle East, and Far East).
The benefits of automation can be extended not only to
road transportation, but to rail transportation as well. Many
large companies, such as Alstom S.A., Bombardier
Transportation, CRRC Transportation, General Electric,
Hitachi Ltd., Kawasaki Heavy Industries, Mitsubishi Heavy
Industries, and Siemens AG., invested substantial funds
into the development and deployment of the autonomous
train (AT) technology. In particular, the AT technology
market was valued at $5.88 billion in 2018 and is
anticipated to reach $15.57 billion by 2026 [10]. The
deployment of ATs is expected to offer the following
advantages to railroad companies [10]: (1) reduction in
accidents and safety improvements; (2) decrease in
operational costs; (3) ATs with a full automation level will
completely eliminate potential risks due to human errors;
(4) reduction in emissions produced; (5) increased capacity
for passenger and freight transport; (6) improved reliability
of rail services; (7) effective communication with
connected vehicles (CVs) and AVs; and others. Along with
advantages, there are some challenges that are associated
with the AT deployment and reaching the full automation
level (e.g., substantial capital costs of automation, required
infrastructure improvements, potential cyber-attacks that
may disturb the AT operations, legal issues, and many
others that will be further discussed in this study).
TABLE I
PERCENTAGE OF RAIL PASSENGERS, FREIGHT, AND RAIL LINES GLOBALLY BY GEOGRAPHICAL LOCATION (2004-2018)
Geographical Location
Rail Passengers (million
passenger-km)
Rail
Passengers (%)
Rail Freight
(million ton-km)
Freight
(%)
Length of Rail
Lines (km)
Length of Rail
Lines (%)
Africa
982,361
2.0
2,315,417
1.5
1,141,683
6.8
America
906,175
1.8
47,590,186
31.8
5,701,525
34.1
Asia and Oceania (Russia and
Turkey excluded)
37,788,381
76.1
56,872,292
38.0
4,492,658
26.9
Russia
2,205,089
4.4
32,280,372
21.5
1,278,660
7.7
Europe (including Turkey)
7,794,595
15.7
10,769,925
7.2
4,096,830
24.5
FIGURE 3. Projected AV growth and benefits.
Although a significant number of survey studies have
been conducted to date aiming to investigate various aspects and challenges associated with the deployment of
CVs and AVs [11-15], very limited survey efforts have

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
4
been geared towards a comprehensive understanding of
advantages and challenges from the AT deployment [16-
18]. Therefore, this study aims to conduct a detailed up-to-
date review of the state-of-the-practice and the state-of-the-
art, aiming to identify the existing trends, technologies,
advancements, and challenges in the development and
deployment of ATs in rail transportation. Based on the
conducted review, this study identifies the main advantages
that can be achieved from the AT deployment. Moreover,
the main challenges from the AT deployment that have to
be addressed in the nearest future are determined as well.
The remainder of this manuscript is organized as follows.
In the second section, the current trends in the railroad
industry with respect to the AT applications are discussed.
The review of state-of-the-art efforts related to ATs is
described in detail in the third section. The state-of-the-art
summary, identified benefits from the AT deployment, and
the associated challenges are presented in the fourth
section. The study conclusions and necessary future
research works are described in the fifth section. A full list
of abbreviations that will be used in this manuscript is
presented in Appendix A.
FIGURE 4. Automation levels for AVs.
II. REVIEW OF THE CURRENT TRENDS IN
AUTONOMOUS TRAIN APPLICATIONS
Advancements in rail technology are quickly taking it to the
next level from partial or no automation to a full
automation level. Highly sophisticated cutting-edge
technologies are being used or planned to be used to
achieve the full automation level in trains. A few of these
technologies that are used in combination are high-speed
internet (5G) technology, infrared cameras, ultrasonic
cameras, dedicated short-range communications,
accelerometers, tachometers, sensors, among others. Based
on the International Association of Public Transport
framework, there are four grades of automation (GoAs) for
trains that include the following (see TABLE II) [19]:
 GoA1: All the major train operations, such as starting,
stopping, door operations, addressing emergency
situations, and sudden diversions, are manual and
involve an onboard driver;
 GoA2: Some of the train operations, such as starting
train, stopping, changing the rail tracks, are automated
but an onboard driver is still needed;

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
5
 GoA3: This level provides autonomous train
operations, but in case of an emergency an onboard
attendant takes control of the train; and
 GoA4: At this level, the train runs fully autonomous
with no onboard driver/attendant (see Fig. 5).
TABLE II
GRADES OF AUTOMATION FOR TRAINS
Grade of Automation
(GoA)
Type of Train Operation
Driver/Attendant
Presence Required
Starting
Train Motion
Stopping Train
Motion
Door Opening
and Closure
Emergency
Situations
GoA1
ATP with Driver
Yes
Driver
Driver
Driver
Driver
GoA2
ATP and ATO with Driver
Yes
Automatic
Automatic
Driver
Driver
GoA3
DTO
Yes
Automatic
Automatic
Attendant
Attendant
GoA4
UTO
No
Automatic
Automatic
Automatic
Automatic
Notes: ATP – automatic train protection; ATO – automatic train operation; DTO – driverless train operation; UTO – unattended train operation.
Although electric vehicles and AVs are constantly
receiving attention and substantial investments, a
significant progress has been done towards the deployment
of ATs with the full automation level. For example, in the
Pilbara region of Western Australia, mining corporation
―Rio Tinto‖ has moved to driverless fully autonomous
operations for its entire rail system (heavy haul) in June
2019 [20], while the first autonomous rail journey occurred
in October 2017. The Rio Tinto’s rail network is recognized
as the world’s first fully autonomous rail network. It is
expected that reaching the level of full automation for ATs
will be more difficult for North America, where multiple
passenger and freight operators have to share the same rail
tracks, trains have various weights and types, and there are
numerous yards and junctions.
Similarly, there are multiple challenges in Europe as
well. However, the Société Nationale des Chemins de Fer
Français (SNCF), the French National Railway,
successfully finished the first test using a locomotive-
hauled AT that was remotely controlled in July 2019. This
test was a part of a larger project aiming to develop
prototypes of autonomous freight and passenger trains by
2022 [20]. New automatic metro lines are being introduced
in Barcelona (Spain), and the Barcelona Metropolitan
Transport authority plans to convert its busiest conventional
metro lines into fully autonomous lines in the nearest
future. The process is going slower than expected mainly
due to complexity of administering such an extensive
project [20]. Similar projects are undertaken towards fully
automated metros in other countries as well (e.g., Brazil,
Canada, Denmark, Germany, Italy, and United Kingdom).
The International Association of Public Transport forecasts
a major acceleration in the development of fully automated
metros in the following years. In particular, the length of
fully automatic metro lines is expected to triple between
2019 and 2023 [20]. Even nowadays, over 40 cities around
the globe have fully operational automated rail lines (e.g.,
Paris, Sydney, Vancouver, São Paulo, and Copenhagen).
Furthermore, there are more than 60 fully automated rail
lines in the world, most of which are located in Asia (e.g.,
Hong Kong, Singapore, Japan, and South Korea).
This section of the manuscript provides a detailed review
of the existing trends, technologies, and advancements
associated with the deployment of ATs in the railroad
industry, including the following: (1) overview of basic AT
features; (2) Internet of Things; (3) artificial intelligence;
(4) dedicated short-range communications; and (5) positive
train control.
FIGURE 5. An example of a fully autonomous passenger train.
A.
OVERVIEW OF BASIC AUTONOMOUS TRAIN
FEATURES
The Railway Technical Research Institute (RTRI, Japan)
addressed the sustainable development goals by introducing
various research activities as a part of the initiative
―RESEARCH 2025‖ and emphasized on the goals of
innovation, industry, and infrastructure for railroads [21].
The main RTRI objective is to develop solutions to
different challenges facing the railroad industry (e.g., global
environmental problems, social burden associated with
aging populations, regional disparity of the economy) in
cooperation with the railroad practitioners, research
institutions, academic institutions, and other relevant

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
6
stakeholders. The deployment of ATs with the full
automation level is one of the strategies to achieve
sustainable development goals. Fig. 6 and Fig. 7 show a
variety of innovative technologies that are generally used
for operations, movement, and maintenance of ATs [21].
FIGURE 6. Basic AT operations.
FIGURE 7. Digitalized maintenance for ATs.
Different technologies are deployed in ATs to provide the
information to moving trains regarding the passengers at
nearby stations, route control and braking patterns, obstacle
detection on the tracks or in the vicinity of tracks, entire
line operations, disaster prevention and maintenance
information, safety and ground equipment control, and the

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
7
information regarding the location of neighboring trains.
ATs also use onboard data measurement devices, which
help in early detection of any abnormalities on tracks and
structures. Multiple-point synchronized-control type elastic
switches, also known as intelligent switches, are deployed
on tracks for detection and collecting some operational
data. Various types of data collected assist with the
development of big data analysis models for improving the
effectiveness of AT operations. The development of new
advanced communication and detection technologies is
expected to facilitate the deployment of ATs and meet the
main sustainable development goals.
FIGURE 8. The number of IoT connected devices globally between 2015
and 2025.
B.
INTERNET OF THINGS
The Internet of Things (IoT) technology uses high-speed
internet and wireless technology to provide seamless
connectivity to various devices, applications, and systems
deployed by ATs. The IoT functionality and performance
directly depend on advanced computing applications, big
data available, machine learning techniques adopted, and
artificial intelligence methods that are used for the analysis
of critical data collected from surrounding CVs, AVs, and
other areas and objects. The data collected and evaluated by
means of IoT allow enhancing the efficiency and safety of
the overall system, where ATs serve as an integral part. The
IoT technology can be effectively used with the Global
Positioning System (GPS) to determine the appropriate
routing choices for ATs and communicate the changing
train location to a central command center and various
infrastructure units.
Such communication is critical, especially at the
locations where ATs can have potential conflicts with
pedestrians or vehicles (e.g., highway-rail grade crossings,
where a highway segment intersects with a railroad
segment at the same elevation and creates a conflicting
point between the arriving trains and vehicles). The number
of IoT connected devices is expected to rapidly increase in
the following years (see Fig. 8), which will provide an
opportunity to the railroad industry and other industries to
benefit from the IoT advantages [22-24]. It is expected that
the IoT technology in the railroad industry will become a
$30 billion market in the next 15 years.
C.
ARTIFICIAL INTELLIGENCE
The artificial intelligence (AI) component of ATs consists
of various interlinked technologies, including the following
[22, 25]: (1) natural language processing technologies; (2)
robotics; (3) machine learning technologies; (4) vision
technologies; (5) speech technologies; (6) planning
technologies; and (7) expert systems. The AI technologies
are used for the simulation of human intelligence and
machine learning. The AI can be applied to solve a specific
decision problem and has huge processing power. The AI-
based technologies rely on mathematical models, software,
and algorithms that can potentially enhance the
performance of a given system. Advanced AI techniques
(e.g. neural networks, genetic algorithms, tabu search,
variable neighborhood search, simulated annealing) can be
applied for optimizing various AT operations (e.g., travel
time optimization, timetable synchronization, AT routing,
intelligent scheduling, AT maintenance procedures,
intelligent inspection).
For example, the SNCF railway has been using a variety
of AI-based techniques for many years (such as prediction
algorithms, real-time data processing, and sequential
machine learning) along with different types of supporting
equipment (i.e., remote sensors for detecting pressure,
vibration, and temperature; devices for automatic alerts,
field and onboard equipment), which allowed substantially
improving the railroad operations. In particular, the SNCF
railway increased its ridership by approximately 50% in the
last 10 years and is capable of successfully running 15,000
trains per day [22].
D.
DEDICATED SHORT-RANGE COMMUNICATION
Dedicated short-range communications (DSRC) can be
viewed as a short-range communication system with a
reliable two-way high-speed radio service that can
be effectively deployed for vehicle-to-vehicle (V2V) and
vehicle-to-infrastructure (V2I) communication [26]. The
DSRC is able to operate continuously in a broadcast-and-
receive mode, which provides situational information
regarding surrounding vehicles. The CVs equipped with the
DSRC technology are able to exchange information at a 5.9
GHz spectrum without having any cellular infrastructure.
The information is exchanged by means of a basic safety
message (BSM) that contains vehicle size, heading,
position, speed, brake status, and steering angle. The BSM
can be frequently broadcast between CVs (e.g., every 100
milliseconds). Unlike many other communication
technologies that are based on one-to-one communications,
the DSRC enables one-to-many communications. The
DSRC technology has been developed to effectively work
in the fast moving environments, where a sender is moving
away from a receiver at speeds that exceed 100 mph. The
DSRC is viewed as one of the most effective alternatives

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
8
for V2V communications due to its unique combination of
attributes (i.e., message broadcasting without a network
connection, high broadcast frequency, trusted and
anonymous communication, and effective functionality in
the fast moving environments).
FIGURE 9. Vehicle communication systems.
Fig. 9 shows two different types of communication
systems. The first communication system deploys the
DSRC technology for road-side unit (RSU), V2V, V2I, and
V2X (which is a combination of V2V and V2I)
communications. One the other hand, the second
communication system allows interactions between
vehicles by means of vehicle-to-network (V2N)
communication. It can be observed that the DSRC enables
more effective interactions between vehicles and
surrounding objects when comparing to basic cellular
technology-based communications [27]. In case of the AT
deployment, the DSRC will use an onboard unit, which will
be installed in the locomotive and will interact with the
radio communication-based RSUs that are installed in the
proximity of rail tracks. The DSRC can provide necessary
warning at highway-rail grade crossings to the surrounding
vehicles regarding an approaching AT in a similar way that
is used to provide warning to AVs on restricted stretches of
highways.
E.
POSITIVE TRAIN CONTROL
The positive train control (PTC) technology has been
developed as a railroad safety system that automatically
slows down a train as soon as it crosses a specified travel
speed or skips a signal [28]. Such a system is mainly
designed to reduce the impacts of potential human errors
and decrease the severity of potential accidents at highway-
rail grade crossings as well as train-to-train accidents. Wi-
Fi, GPS, and high-band radio transmission are used by the
PTC system to identify the location, direction, and speed of
the approaching trains. The train crew can be provided a
notification if there are any potential issues on rail tracks.
Furthermore, the approaching train can be controlled
remotely in case there is no timely response from the train
crew. The PTC system has been primarily deployed at
freight railroads but has the potential for passenger railroad
applications with some adjustments in terms of equipment
and software. Fig. 10 shows different components that are
directly used by the PTC system. The Advanced Civil
Speed Enforcement System (ACSES) is one of the main
PTC components and represents the locomotive cab
signaling system. Advanced signaling and communication
capabilities provided by the PTC system are expected to
improve safety of roadway users and train crew at the
railroad networks with conventional trains and ATs as well.
III. REVIEW OF THE RELEVANT STATE-OF-THE-ART
EFFORTS
Many state-of-the-art research efforts have been conducted
to date to investigate various aspects of ATs, such as
current trends, stages of development, use of technologies,
safety and reliability issues, level of automation, challenges
in the deployment, and others. A detailed up-to-date review
of the relevant literature was conducted as a part of this
study following the content analysis method, which is
considered as a well-established method for systematic
reviews of the literature [29]. A thorough literature search
was performed via the databases supported by the top
scientific publishers (e.g., Elsevier, IEEE, Springer, SAGE,
Emerald, etc.) to identify the studies that are the most
relevant to the theme of the present survey using the
following keywords and phrases: ―autonomous trains‖,
―freight train automation‖, ―passenger train automation‖,
―autonomous train technologies‖, ―rail automation
advantages‖, ―rail automation challenges‖, ―autonomous
train safety‖, ―autonomous train reliability‖, and ―rail
automation perception‖.
The search engines of the considered publishers
identified many thousands of articles. However, only a
small portion of those articles were dedicated to
autonomous rail transportation, while the majority of the
articles strictly concentrated on autonomous road
transportation (i.e., focusing on CVs and AVs without
providing any connection to ATs). The collected studies on
the AT deployment were classified into the following
categories: (1) CAV technologies and their applications for
ATs; (2) general issues associated with ATs; (3) user
perception and outlook for ATs; (4) design and
technologies for ATs; and (5) ATs and highway-rail grade
crossings. The following sections of the manuscript present
a description of the collected studies.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
9
FIGURE 10. Design of a PTC system.
A.
CONNECTED AND AUTONOMOUS VEHICLE
TECHNOLOGIES AND THEIR APPLICATIONS FOR
AUTONOMOUS TRAINS
As a result of the conducted literature search, many studies
dealing with the CAV technologies were identified, and
only some of them discussed how these technologies could
potentially influence the AT deployment and operations.
For example, Bagloee et al. [30] analyzed various issues
and opportunities regarding the transportation policies and
regulations associated with the CAV technologies. The
study highlighted that the deployment of CAV technology
is expected to reduce the cost of transportation and increase
the accessibility. An essential part of the CAV technology
was found to be its ability to communicate with moving
vehicles and infrastructure, as well as the possibility of an
efficient optimized intelligent routing system. This
technology would be important for the vehicles passing
through highway-rail grade crossings and interacting with
crossing infrastructure as well as the trains approaching the
crossings.
Krasniqi and Hajirizi [31] investigated the current CAV
market and technology trends from the preliminary stage to
fully driverless vehicles. The IoT technology was found to
have a potential to completely alter the automobile market,
which could provide a major thrust to IoT in the
autonomous technology market. The study listed the major
challenges of the autonomous industry, various advantages
and disadvantages, and issues related to its deployment.
The major focus area of the study was related to the
industry thrust of the CAV technology rather than the
academic research. The following major CAV-related
issues were identified: (1) lack of software that are fail-
proof; (2) lack of dynamic maps that can update quickly
and provide detailed information regarding street views;
and (3) sensors that can detect unanticipated conditions.
The study concluded that the DSRC and 5G technologies
would be the most suited for improving safety and vehicle
communication with trains and surrounding infrastructure.
Crains [32] presented a perspective on the future of
autonomous technology in all modes of transportation (not
only AVs but ATs, autonomous aircrafts, autonomous
vessels, etc.). The study highlighted the Copenhagen metro
and Rio Tinto in Western Australia as some of the
successful examples of using autonomous passenger
locomotives and heavy haul freight locomotives,
respectively. It is predicted that the train technologies, such
as hyperloop, may advance to a fully autonomous mode,
especially for passenger transportation. A number of
challenges with the AT deployment were pointed out (e.g.,
introduction of the PTC system in the United States (U.S.)
for freight railroads only; disagreement of railroad unions
to adaptation of one-person train crews), which can slow
down rail automation if not addressed properly. The study
also discussed some technologies that are used in AVs for
different purposes, which are comparable to similar
technologies that are used in ATs (see TABLE III).
Bucsky [33] studied the existing condition of freight traffic

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
10
and CAV applications in freight transportation. It was
highlighted that the level of automation would vary from
one mode of freight transportation to another. The highway
vehicles are expected to achieve the full automation level
with the development of new adaptive digital technologies.
The deployment of autonomous trucks could create
substantial social costs, as truck drivers would lose their
jobs. The study highlighted that the railroad industry would
have more developments towards the full automation level.
However, such developments would require significant
investments. It was indicated that automation would make
the road transport more preferential as compared to the
other modes, which might lead to some environmental
problems.
TABLE III
COMPARISON OF AUTONOMOUS TECHNOLOGIES USED IN AVs AND ATs
a/a
Autonomous Vehicles
Autonomous Trains
Utility of Technology
1
Cameras
Infrared Cameras
Read signage, traffic control devices, lane markings, surrounding environment, etc.
2
Laser Illuminating
Detection and Ranging
(LIDAR)
Laser Illuminating
Detection and Ranging
(LIDAR)
Create 3D maps and help detecting potential hazards. Can determine the distance and
object profile by bouncing the laser beam off the object surface.
3
Radar
Radar
Accurately measures the speed of nearby vehicles and trains in real time that cannot be
adequately achieved by using LIDAR.
4
Sensors
Ultrasonic Sensors
Sensors perform the role of a self-monitoring device to ensure that a vehicle/train is not
speeding and monitor the overall vehicle/train functionality. Also, they perform the role
of object detection.
5
Dedicated Short-Range
Communications
(DSRC)
Dedicated Short-Range
Communications
(DSRC)
The DSRC is a short-range communication system with a reliable two-way high-speed
radio service that can be effectively deployed for V2V and V2I communication. It can
provide warning at highway-rail grade crossings regarding an approaching AT.
6
Stereo Video
Stereo Video
The stereo video uses two cameras to capture a 3D-environemnt that forms the basis for
various assistance systems in the vehicle/train. It helps in measuring the depth
accurately.
7
Human-Machine
Interface (HMI)
N/A
It is a combination of systems inside the vehicle, which includes panels and controls for
the interaction between the vehicle and its occupants.
8
Domain Controller
Domain Controller
This is the main ―brain‖ of the autonomous driving system that controls the signals and
information from LIDAR, sensors, cameras, etc. and determines necessary actions
accordingly.
9
Motion Control
Systems, Actuators,
Mechatronic Units
Motion Control
Systems, Actuators,
Mechatronic Units
They work in combination with other technologies for execution of different actions that
were received from the domain controller.
10
N/A
Positive Train Control
(PTC)
It is a GPS-based technology that is used to stop the train, avoid collision, and any
unwarranted train movements.
The Governors Highway Safety Association (GHSA)
[34] highlighted that the deployment of autonomous driving
systems might create new and unanticipated safety issues
(e.g., compliance with traffic laws, compliance with local
practices, decisions during emergency situations,
recognition, reaction, operations after detection of certain
system failures, system security, and others). These issues
could occur despite recent technical advancements, such as
forward collision warning, obstacle detection, curve speed
warning, and high speed alerts [35]. It was recommended
that partially automated driving systems should be
controlled by a licensed human operator, who can take
control of a vehicle (or a train) when needed.
Elliot et al. [36] conducted a detailed review study,
primarily focusing on recent advancements in CAV
technologies. The study highlighted the importance of the
DSRC and 5G technologies for effective communication
between CAVs, which can be also used for communication
between CAVs and trains. It was indicated that the
integration of various components into one platform
without causing any potential issues for security, reliability,
and conflicts in operations would be an important next step
in the CAV deployment. Another challenge might be
associated with the integration of pedestrian and collision
avoidance system into the intersection navigation control.
The Congressional Research Service (CRS) [27]
discussed the issues related to the deployment and testing of
CAVs. Some of the major issues concerning the
deployment of CAVs were found to be data security and
protection against any intrusion in the onboard computer
system. With the use of CAV technology, many functions
currently being performed by the driver will be performed
automatically. The interaction of CAVs with the existing
infrastructure, ATs, and non-CAVs will generate a huge
amount of data related to vehicle location, train location,
stability, movements, and others. Any unauthorized access
may lead to safety issues not only for the vehicle and for
the train but for others as well. Fig. 11 shows various
vulnerable points in a CAV that could be potentially
targeted by hackers [27]. Disabled DSRC systems due to
the intrusion of hackers would lead to the loss of
communication with the approaching ATs that might
further create some safety issues. In order to prevent the
intrusion of hackers, remote software updates are required
for CAVs and ATs on a regular basis. The manufacturers of
CAVs and ATs are also mandated to comply with a set of
cybersecurity principles and report any cybersecurity
threats, incidents, and violations after their occurrence. The
data collected from CAVs and ATs could be of interest to
other entities as well (e.g., law enforcement agencies,

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
11
insurance companies, urban planners, first responders in case of an accident occurrence, etc.).
FIGURE 11. Vulnerable CAV points that can be subject to an unauthorized access.
B.
GENERAL ISSUES ASSOCIATED WITH
AUTONOMOUS TRAINS
Several studies concentrated on various issues due to the
AT deployment. For example, Gebauer and Pree [37]
discussed some of the conceptual, technical, and legal
challenges of driverless trains on the existing railroads. The
concept of ―trainlets‖ was proposed to improve the
attractiveness of ATs and make them a more competitive
transportation mode. In particular, the entire AT could be
divided into many smaller ―trainlets‖, which have the size
of a standard automobile. Passengers having the same
destination could be placed in the same ―trainlet‖ and travel
directly there with a limited number of stops, which could
reduce the total delay due to unnecessary stops. The study
discussed the autoBAHN project that was launched to
validate the technical feasibility of fully autonomous
―trainlets‖, determine economically promising scenarios,
demonstrate some applications of ―trainlets‖, and address
some of the associated legal issues. A set of simulation
experiments were conducted for a short track with a length
of 13 km, 13 stops, and an average of 400,000 passengers
per year. It was found that ―trainlets‖ were more effective
than regular trains, which operate on their given schedules,
even with a fairly low occupancy of ―trainlets‖ (i.e.,
approximately 1.3 passengers per ―trainlet‖). The
developed simulation model could be used to justify the
economic feasibility of deploying ―trainlets‖.
Gebauer et al. [38] and Gebauer et al. [39] also discussed
various aspects of the autoBAHN project (i.e., technical,
conceptual, and legal). It was mentioned that ATs would
have less legal issues that are associated with accidents,
when comparing to roadway vehicles, as the access to rail
tracks is generally more restricted and managed by the train
control systems. Moreover, the economic viability of ATs
would depend on how the automation concept would be
implemented. In particular, the use of the existing railroad
infrastructure with the minimum required alterations would
be favorable for the development and deployment of ATs.
A new railroad system with ATs should be at least as safe
as the existing railroad system in order to meet the
regulatory requirements. The autoBAHN project is
expected to facilitate the AT deployment on railroads in
Europe. Modern technological features (e.g., obstacle
recognition system, GPS-based train control system, audio
and video communication features) are anticipated to
improve the operations of ―trainlets‖. Given the existing
interest of manufacturers, railroad companies, and other
relevant stakeholders, the autoBAHN project might be
implemented in the near future along the major rail lines.
Nevertheless, some political and legal requirements must be
satisfied before wide implementation of ATs (e.g., the need
for railroad security and safety standards with respect to the
AT testing processes).

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
12
Dordal and Avila [40] presented an intelligent approach,
which was based on software agents, to coordinate trains
along a single railroad track. The main objective was to
improve the utilization of a railroad track and reduce the
associated environmental impacts. The behavior of agents
was based on certain operational attributes (e.g., specialized
rules of conduction, train position, direction, target time to
the final destination, etc.). The results showed that the
proposed software agent-based system could reduce the
total journey time by 22.5%, when comparing to the
traditional method of conduction. Furthermore, the fuel
consumption was reduced by more than 25%, which could
further improve the environmental sustainability of rail
transportation and decrease the emissions produced. Hazan
et al. [41] studied the impacts of AV deployment on rail
transportation. The study pointed out that the introduction
of AVs to the market would have many advantages for
passengers (e.g., reduced road congestion, increased safety,
and more productive travel time). As a result of conducted
survey, it was found that about 50% of respondents (out of
5,500 consumers from ten different countries) would
consider buying AVs. More importantly, at least 40% of the
current train passengers would shift to using AVs. The
introduction of ATs with innovative technological features
might be able to keep rail transportation competitive and
offset the AV benefits.
Powell et al. [42] aimed to assess potential benefits and
challenges of the AT deployment for the Tyne and Wear
Metro (United Kingdom – U.K.) by means of simulation.
The results from the conducted simulation analysis
demonstrated that a substantial capacity increase could be
achieved from the AT deployment when implemented with
signaling upgrades. However, low adhesion conditions
were found to be one of the obstacles. Moreover, additional
infrastructure upgrades would be necessary to transition
towards full automation. The most significant obstacle was
found to be the presence of shared operations at the existing
rail lines. Wang et al. [16] conducted a survey of driverless
train operations in urban transit rail systems. Along with
potential advantages (e.g., lower operational costs,
increased capacity, increased flexibility, improved
reliability, energy efficiency), a wide range of AT
deployment challenges were outlined as well, including
safety issues, train control technology issues,
communication issues, issues associated with platform
screen doors and guideway intrusions, terminal design
challenges, as well as detection and management of
emergency situations. The study pointed out that a
systematic safety assessment framework is needed to
standardize the AT operations, reduce potential risks, and
enhance the overall reliability.
Trentesaux et al. [43] pointed out the growth of ATs,
since they have become an important component of
competitiveness for many fleet operators. Some of the
major benefits of ATs were highlighted, including the
following: (1) efficient utilization of the existing
infrastructure; (2) effective energy consumption; (3)
enhanced quality of service (e.g., better management of
arrivals and departures of trains during unpredicted peak
traffic demand); (4) improved capabilities in terms of
perception; (5) better control than human drivers; and (6)
accurate detection of hazardous situations. Fig. 12 shows an
example framework for nominal AT functioning under
planned and unplanned operational conditions [43]. It can
be observed that the original operational plan could be
adjusted by ATs as needed depending on the external
circumstances. Apart from various benefits that ATs might
be able to provide, various risks and issues that are
associated with the AT deployment were pointed out as
well: (1) design and operational risks in ATs; (2) issues
regarding decision, information, and learning process; (3)
safety, ethics, and norms-related issues; (4) fleet
interoperability and coordination related issues; (5) human
skills, acceptance, and interaction-related issues; and (6)
design methodology, deployment, and process related
issues. The study concluded that the deployment of ATs
would significantly change the railroads. However, the
success of ATs will depend on the participation of various
stakeholders, such as manufacturers, fleet operators,
infrastructure developers, etc.
Wardrop [44] discussed the rationale of developing ATs
as well as the challenges that were associated with the
deployment of heavy haul freight trains in remote areas.
Mining in remote areas is generally expensive, as it is quite
difficult to find the required resources. Therefore, the
automation of mining process and rail transportation gained
popularity among mining companies. Some major issues of
deploying trains in remote areas were underlined, including
the following: (1) wild animals could appear on railroad
tracks, since the tracks are unfenced; (2) highway-rail grade
crossings are generally equipped with passive protection
devices, which may not lead to an adequate safety level in
some instances; (3) detection and mitigation of accidents
when trains collide, derail, or break down; and (4) timely
response to different incidents that may occur on railroad
tracks and stop the AT movement. The study concluded
that the combination of onboard driving control systems
and remote train dispatching should improve the control of
train flows, timekeeping of individual freight trains, and
enhance the utilization of rail lines. It was also indicated
that the practices that were used for remote heavy haul
freight ATs could be implemented to suburban passenger
ATs as well.
Antolini [45] focused on the project conducted by the
French National Railways (SNCF) company, aiming to
develop the prototypes of ATs with partial and full
automation levels. The first test for a remotely controlled
locomotive-hauled AT was completed by the SNCF in July
2019. The overall goal of the project is to design ATs with
a full degree of automation for freight and passenger
transport by 2022. The study highlighted that the main
challenge of the AT deployment in Europe would be the
introduction of ATs on the railroad system, where multiple
passenger and freight operators share the same railroad

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
13
tracks. Furthermore, the trains might have different physical
characteristics (i.e., train types and train weights), which
could also impose additional operational challenges. French
railroad networks have numerous yards and junctions that
could potentially impose difficulties for the AT
deployment. Despite these challenges, the SNCF remains
optimistic and aims to effectively achieve the project
objectives. The project with the expected overall cost of 57
million EURO is anticipated to improve punctuality of
operations, provide larger capacity, and decrease the
environmental impacts.
FIGURE 12. Nominal functioning of an AT.
Kemmeter [46] underlined some of the potential
challenges associated with the AT deployment. Certain
CAV technologies may not be applicable for ATs (e.g.,
LIDAR cannot be used to measure the distance to the next
train, as the distance between consecutive trains is
significant when they operate at a full or close to the full
speed). Furthermore, train detection requires installation of
cables along the rail tracks as well as a lot of online
equipment. Such train detection equipment is generally
costly and requires regular maintenance. Muller [47]
evaluated the organizational hindrances from the
deployment of freight ATs. The study evaluated four
conceptual impacts on innovation activities within
advanced markets, including the following: (1) nested
architecture and hierarchy of networks; (2) dependency on
the path in technological prototypes; (3) dynamics of
organization; and (4) technological standoff. It was pointed
out that the aforementioned impacts could directly
influence the innovation activities. The conducted analysis
demonstrated that the economical processes could impose
substantial limitations on the innovation activities for the
general rail transport operations and freight AT operations.
The study concluded that the strategic innovation policies
should address the challenges that are associated with the
AT deployment.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
14
Pattinson et al. [48] highlighted that the future AVs will
have to co-exist with the vehicles that are driven by
humans. It would be critical to develop the appropriate
solutions to determine legal responsibilities for vehicles and
drivers in case of accidents. The AV users should
understand all the risks involved when deciding to drive
such vehicles. Similar legal issues have to be addressed
when deploying ATs and human-driven trains on the
existing rail lines. Wang et al. [49] focused on the safety
issues associated with CAVs. Drivers often take control of
AVs when they think it is necessary, which creates a
―disagreement‖. Based on the reported data, the rate of such
disagreements can be fairly high with up to 3 disagreements
per mile for different manufacturers. Such disagreements
may further lead to accidents. The study also showed that
the majority of accidents are caused by other roadway users
– not CAVs. Similar studies should be administered for
automated rail systems (especially, for the rail lines that
have trains with the grades of automation GoA2 and GoA3,
where there may be a potential disagreement between the
partially automated train and the driver or the attendant).
Othman [50] indicated that the COVID-19 pandemic
made substantial impacts on public transit services around
the world, including rail transportation. The ridership on the
U.S. rail and bus systems reduced by 79% in Washington,
D.C., 83% in Boston, and 74% in New York. Similarly, the
transit ridership decreased by more than 80% in Montreal
and Toronto (Canada). The future rail transit systems,
including autonomous rail transit systems, should have
additional protective measures against the spread of
airborne diseases (e.g., advanced air circulation, ultraviolet
light disinfection). Such measures will help preventing the
impacts of airborne diseases on rail transit systems and
improve safety of passengers.
The AT deployment is expected to reduce the train crew
size, as the ATs with a full automation level will not require
any onboard personnel. A number of studies discussed the
issues associated with the train crew size reduction.
Karvonen et al. [51] underlined the importance of train
drivers in the metro operations and potential difficulties in
transitioning towards fully automated metro lines. Based on
the data collected from the Helsinki Metro (Finland), it was
found that train drivers actually play a critical role (i.e.,
anticipation, observation, interpretation, and reaction to
events), which may not seem apparent in some instances.
Moreover, train drivers are viewed as a link between
different actors involved in different metro operations. It
was concluded that a lack of understanding of the train
driver roles may lead to safety and quality of service issues
throughout the AT deployment. Cohen et al. [52] used
semi-structured interviews and questionnaires, distributed
among the automated metro line operators, to determine
how automation influenced staffing, costs, reliability, and
service capacity. The results, which were collected from 23
metro lines, indicated that the AT deployment could
decrease the train crew size by 30-70% (which can be
negatively viewed by the public). Moreover, the survey
indicated that the capital costs of the automated metro lines
were fairly high, but the internal rates of return were
promising in some cases. The findings showed that the AT
deployment could improve the metro efficiency and
productivity. However, additional data would be needed to
draw more accurate outcomes.
Cassauwers [53] discussed the employment issues due to
the AT deployment. It was highlighted that the introduction
of automated systems in passenger and freight rail
transportation could lead to the train crew size reduction,
which might cause a large number of strikes by railroad
unions. Many experts and professionals still have concerns
regarding the AT technology performance in emergency
situations. The study pointed out that the re-orientation of
employees (e.g., transition to customer service or to non-
automated rail lines) would be a promising solution rather
than layoffs or salary cuts.
C.
USER PERCEPTION AND OUTLOOK FOR
AUTONOMOUS TRAINS
Only a few studies investigated the user perception and
outlook for ATs. Fraszczyk et al. [54] indicated an
increasing trend for the AT deployment and underlined the
lack of studies focusing on public perception of substantial
changes in rail transport associated with ATs. The study
specifically concentrated on public perception for the
automated metro systems with unattended train operations.
A survey was conducted as a part of the study, which
involved a total of 50 individuals from 10 different
countries. Approximately 75% of the respondents were
below 30 years of age. Based on the gender distribution,
72% of the respondents were male and 28% were female.
The results of the study showed that 72% of male
participants and 93% of female participants approved
having a ―fake‖ driver room on ATs. The majority of
survey participants indicated no concerns associated with
the AT maintenance issues and the train design itself.
However, staff communication and technical failures were
the main two issues that were raised by some of the survey
participants. Moreover, most of the survey participants did
not raise any employment concerns due to the AT
implementation (i.e., train drivers will not be needed for
ATs).
Fraszczyk and Mulley [55] studied the user perception
and attitude towards ATs in Sydney (Australia). It was
pointed out that the AT deployment project in Sydney
Metro was technically sound and well-planned, despite the
fact that it was introduced late in the planning phase. A
survey among 300 participants was performed to meet the
study objectives. An almost even distribution of female and
male participants was achieved. The study results indicated
that many train users and non-users rated the presence of a
train driver as important. Moreover, the majority of survey
participants were concerned with the safety aspects of the
AT operations. The users identified ―reduced ticket price‖
and ―increased train frequency‖ as the main benefits from
the AT deployment. On the other hand, ―energy recovery

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
15
and sustainability‖ were defined at the least important
benefits. Based on the survey results, approximately 50% of
the existing users indicated that they would use the fully
automated metro in the future, and about 35% of the
existing users were not sure. In the meantime,
approximately 40% of non-users indicated that they would
not be willing to use the fully automated metro in Sydney.
Pakusch and Bossauer [56] highlighted that autonomous
public transportation has a lot of benefits and can enable
sustainable mobility. The study focused on the user
acceptance with respect to the technological innovations
and developments in autonomous public transportation. A
survey among 201 participants was conducted to achieve
the study goals. The survey was advertised via different
online platforms and social networks in Germany. The age
of survey participants varied from 18 to 81 years, with an
average age of 26.2 years. A total of 49.3% of the survey
participants were female. Most of the survey participants
were students due to a number of reasons (e.g., reduced cost
of tickets for students, well-developed urban transport in
large metropolitan areas). The study results indicated that
many users were already familiar with autonomous driving
and were willing to use autonomous public transportation in
the following years. It was highlighted that the existing
policies should be modified to allow the users accessing
autonomous public transportation even during the test
phases, so they could become more familiar with new
technologies and have positive experience in the future. The
accessibility to autonomous public transportation would
further lead to sustainable mobility behavior of users.
Henne et al. [57] discussed the challenges that are
associated with the perception of AI-based applications for
autonomous systems. It was underlined that the user
perception could substantially affect the deployment of
autonomous systems. In particular, AVs and ATs should
drive safely in various environments and complex situations
in order to have a positive user perception. One of the main
AI advantages consists in the fact that the AI techniques are
capable of abstracting from distinct learned scenarios to
effectively solve the present tasks. The study identified a
set of approaches that could be potentially used to model
the uncertainty in AI-based perception (e.g., uncertainty in
model parameters, uncertainty in data). The following
approaches were discussed: (1) calibration of the model
outputs; (2) detection of out-of-distribution inputs; (3)
application of Bayesian Neural Networks; and (4)
application of deep ensembles. A combination of methods
for assessing the perception uncertainty and dynamic
dependability management was found to be a promising
solution to effectively address the challenge of unreliable
perception of the AI-based applications for autonomous
systems.
D.
DESIGN AND TECHNOLOGIES FOR AUTONOMOUS
TRAINS
A number of studies proposed innovative concepts and
models that could be used for the future AT design. Bock
and Bikker [58] presented a new operational concept for
rail services, which is inspired by driving on demand. The
proposed concept assumed that the train wagons were no
longer physically coupled to each other. Furthermore, each
wagon had its own computer system along with propulsion.
Therefore, each wagon would become an intelligent unit,
which can be considered as an AT module prototype.
Electronic data transition would allow controlling these
trains, so they could drive close to each other with the
minimal distance. Each train module was able to join the
formation of modules at rail junctions or leave the
formation (depending on the final destination). The
proposed operational concept was expected to make rail
transportation more competitive as compared to road
transportation. Haxthausen and Peleska [59] introduced the
notion of a controlled distributed system for railroads and
presented the specification and authentication of the main
algorithm for safe distributed control. The required safety
levels were derived based on the abstract vision that could
be easily verified with respect to their completeness and
soundness. The overall complexity was decreased by
dividing the system model into two sub-models, including a
controller model and a domain model. It was highlighted
that the proposed concept could be used for an automatic
train control without the presence of a train driver (i.e., AT
applications in rail transportation).
Matsumoto et al. [60] discussed the level of reliability,
safety, and capacity that is needed for the train control
system at busy railroad sections. The study proposed a
decentralized autonomous train control (D-ATC) system
that can fulfill the existing needs. The proposed D-ATC
system had the following main features: (1) train detection
on rail tracks; (2) dissemination of the stop point
information; (3) detection of a train position; and (4)
control of braking. One of the highlighted challenges was to
configure the available technologies in a way that would
allow the current train control system to co-exist with the
newly introduced system. Fig. 13 shows a comprehensive
framework that could be used for introducing a new train
control system into the existing one [60]. Such a framework
ensures the co-existence of new and conventional train
control systems and, more importantly, enables continuous
train operations without disruptions. The study pointed out
that it would not be feasible to remove the existing
automated train control (ATC) devices at the considered
railroad sections and install new D-ATC devices in a one
train interval (e.g., the last train  the first train). The latter
can be justified by the fact that such a procedure would be
very expensive, and the replacement time may be
insufficient at certain railroad sections. The proposed
technology was deployed for the Yamanote and Keihin
Tohoku lines (Japan).

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
16
FIGURE 13. A framework for co-existence of train control systems.
Matsumoto [61] discussed various train control systems
that were deployed in Japan. The original ATC systems
were applied at the rail lines to ensure safe and fast rail
transportation. However, the conventional ATC systems
were not effective at busy rail lines. Busy rail lines were
generally equipped with the D-ATC systems. The study
pointed out that the newly designed information
technologies allow trains detecting their own positions and
transmitting this information to other trains in the vicinity.
The train control system, which directly relies on the newly
designed information technologies, had been deployed at
the Senseki line of the Tohoku District (Japan). The results
of pilot tests confirmed the effectiveness of the new train
control system. Castells et al. [62] discussed different
advantages from the AT deployment at metro lines,
including increased capacity, flexibility, viable service
during off-peak hours, greater efficiency and safety, and
cost-effectiveness. It was highlighted that the AT signaling
and control systems allow shorter headways between
consecutive trains, which increases the rail network
utilization. The AT deployment allows better management
of the supply and demand (e.g., unused ATs could be easily
removed from the network during off-peak hours). ATs also
have the ―anti-bunching‖ feature that prevents them from
getting stuck between consecutive metro stations. Different
examples of automated metro systems were presented (i.e.,
Vancouver, São Paulo, Paris, and Copenhagen).
Marrone et al. [63] pointed out the importance of
verification process of different AT applications in
automated metro systems. The study indicated that the
Combined Model-Based Analysis and Testing of
Embedded Systems (MBAT) project conducted in the
European Union would help improving the verification
process. The procedures adopted for the verification
process in automated metro systems could be extended to
other complex systems as well. Mohammed et al. [64]
developed a microcontroller-based prototype for ATs. The
proposed AT prototype was able to perform a set of basic
operations, including the following: (1) travel via a pre-
defined path between a set of stations; (2) sensing the
station arrival; (3) proper stopping at stations; and (4)
display of synchronized messages regarding the train arrival
at a given station. Furthermore, the AT prototype was
designed to produce alarm signals to prevent potential
safety issues. The study concluded that the proposed
microcontroller-based AT prototype could serve as a
foundation for more sophisticated control systems.
Balasubramanian [65] discussed various IoT processors
that could be utilized to deploy ATs for a long-distance
travel. A major emphasis was given to the following
aspects: (1) safe and uninterrupted communication between
different components of the train during movements; (2) the
auto record of errors: (3) use of emerging technology (i.e.,
IoT); (4) steady implementation of various solutions; and
(5) availability of backup options to avoid the dependency

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
17
on a single component. Fig. 14 shows an example of
communication paths for ATs throughout their long-
distance travel. It can be observed that the AT
communication paths involve many different components,
including manned level crossings (MLCs), track change
points (TCPs), control center, signaling cabins, satellite
systems, and others [65]. A successful implementation of
the IoT technologies is heavily dependent on the
availability of high-speed internet. The proposed concept of
long-distance ATs is expected to have a variety of
advantages, including the following: (1) improved safety
level for the AT itself and surrounding trains as well; (2)
improved safety performance at highway-rail grade
crossings and bridges; (3) provide more effective service to
the existing train users; and (4) provide more flexibility to
railroad authorities in terms of system improvement
projects.
Fraga-Lamas et al. [17] highlighted that the railroad
industry could benefit from exploiting various opportunities
that are offered by the Industrial Internet of Things (IIoT),
which is an important component of the paradigm called
―Internet of Trains‖. The study presented a comprehensive
review with regards to the evolution of rail communication
technologies, showing how the rail technology
specifications, requirements, and recommendations had
been evolving over the past years. The advantages of
deploying the latest broadband communication technologies
(e.g., 5G technologies, Long-Term Evolution [LTE]
technologies) were explained in detail. Moreover, the
conducted study presented a set of scenarios and
architectures for the railroad industry, where it could attain
better commercial IIoT capabilities. The latest
advancements in smart infrastructure, asset monitoring,
predictive maintenance, video surveillance, safety
assurance systems, train control systems, signal systems,
energy efficiency, and cybersecurity systems were
discussed as well. It was concluded that the IIoT and
Internet of Trains still face a large variety of issues (e.g.,
interoperability, standardization, scalability, and
cybersecurity) that have to be effectively addressed by
researchers and practitioners in the following years. New
types of technology should be tuned for specific railroad
environments to ensure their adequate implementation.
Fig. 15 shows a large variety of IIoT-enabled services
that are specifically related to rail transportation. It can be
observed that the rail IIoT-enabled services cover a wide
range of different areas, including the following [17]: (1)
information (passenger information systems and freight
information systems); (2) train control systems; (3)
predictive maintenance; (4) smart infrastructure; and (5)
energy efficiency. Furthermore, based on their
functionality, the IIoT-enabled technologies for rail
applications can be categorized in different groups that
include, but are not limited to, the following [17]: image
processing and computer vision, algorithms and methods,
sensors, modeling and simulation, communication systems,
computing, big data and data analytics, and artificial
intelligence (see Fig. 16). The deployment of the IIoT-
enabled technologies is expected to enhance rail operations
and maintenance. Furthermore, these technologies will
improve the quality of rail services for both passenger and
freight components. Researchers and practitioners
anticipate that the IIoT will revolutionize the train
operations and make rail transportation a more competitive
mode.
FIGURE 14. Communication paths for ATs.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
18
FIGURE 15. IIoT-enabled railroad specific services.
Romano et al. [66] indicated that several AT metro
systems had been introduced over the past years in certain
countries. However, the demand for rail transportation is
constantly changing, and it is instrumental to design
flexible and efficient ATs, which could support the
existing requirements without affecting the overall
performance and safety. It was highlighted that the new
generation ATs should have the appropriate types of
technology that could achieve operational cost
optimization, energy reduction targets, minimization of
risks associated with human errors, and high system
performance. The modern ATs should have a variety of
features that would allow meeting rigorous customer
requirements. Bharathi et al. [67] presented a technology
for automatic train control and operations that could be
used for metro train movements. The train was assumed
to have a controller enabling the train automatically move
from one metro station to another (i.e., an AT prototype).
Upon arrival at a metro station, the train could stop
automatically by applying the appropriate sensors. The
proposed train system was also capable of counting the
number of passengers walking in and out of the train
doors, which could be further used for estimating the train
capacity utilization. The train was also equipped with a
buzzer to alert the surrounding passengers regarding the
closing doors. Such an automatic train system could be
effective as it eliminates potential human errors.
Lagay and Abdell et al. [68] discussed various features
of the project conducted by the SNCF railroad company,
aiming to deploy ATs with a full automation level on
French railroads. The AT control system was assumed to
have three layers, including the AT protection layer, AT
operation layer, and AT supervisor layer. A variety of AT
technologies and systems were described (e.g., wireless
communication, cybersecurity subsystem, train
positioning, sensor fusion and data processing, AI,
obstacle detection, signaling recognition, and monitoring
subsystems). In order to effectively deploy ATs, the
SNCF planned to work with different industrial and
service sectors, as well as low-cost solutions and high-
performance solutions. Arup [69] provided a detailed
review of new trends in rail transportation that would
further define its development in the following years. The
study presented the examples of ATs deployed for
passenger and freight transportation. It was highlighted
that the AT technologies allow optimizing the running
time, increasing the average speed, and operations in a
close proximity to other trains. The Dubai Metro was
described as an example of the largest driverless metro
network in the world (with approximately 75 rail km).
The AT deployment also allows effectively addressing
the growing demand for rail transportation (e.g., São
Paulo’s Metro Line 4 is the only driverless metro line that
handles 700K passengers per day on a 8-km stretch).
Kimiagar [22] discussed the emerging trends in the
technology used for rail transit, such as AI, train control
systems, dynamic route optimization, predictive
maintenance, and simulation modeling. Various AI
braches that could be applied in rail transit were described
(e.g., machine learning, image and obstacle recognition,
robotics, and planning). The IoT applications in rail
transportation are projected to exponentially increase in

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
19
the following years. The study presented different train
control systems that are responsible for various functions,
including the train arrival time estimation, timetable
development, ride comfort, safety aspects, interlocking,
train position determination, and safe braking. The steps
required for the transition from conventional trains to
ATs were pointed out as well. Pickering et al. [70]
underlined the need for innovative solutions that could be
used to effectively address the increasing demand for rail
transport in the U.K. The study presented a simulation
model that emulated the rail service between Coventry
and Birmingham, which are both located in the U.K. The
results from the conducted simulation analysis
demonstrated that the AT deployment could substantially
increase the rail network capacity, as consecutive trains
could be operated at a smaller distance (taking into
consideration the ―safe‖ separation distance between
consecutive trains). The study highlighted that a smaller
distance between consecutive ATs could be achieved via
the implementation of advanced technologies (i.e.,
accurate train positioning, V2V communication,
predictive braking, fast and reliable switches).
FIGURE 16. IIoT-enabled technologies for rail transportation.
A few studies concentrated their efforts on railroad
signaling system modeling and improvements. Kunifuji et
al. [71] indicated that the efficiency of the existing railroad
signaling system was quite low (e.g., the time allocated to
certain activities along rail tracks was not sufficient due to
low flexibility of the existing signaling system). The study
proposed an alternative railroad signaling system that was
based on the network and autonomous decentralized
technology. It was highlighted that the developed system
would improve not only the flexibility of signaling but the

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
20
overall safety and reliability as well. The main challenge
with the deployment of new railroad signaling system
consisted in the fact that re-modeling of the control logics
might be required for different types of signal devices. Harb
et al. [72] introduced the FRSign dataset, which can be
viewed as a large-scale dataset for vision-based railroad
traffic light recognition and detection. The study
highlighted a lack of open source data for the AT
deployment. The recordings for the FRSign dataset were
made on selected operational trains in France. An
illustrative dataset with more than 100,000 images, which
covered a total of six types of French railroad traffic lights
and their possible combinations, was released as an open
source. The developed dataset can be used for automatic
classification of railroad signaling panels by state and by
type. Furthermore, the developed dataset could assist to
facilitate the AT deployment.
Some of the previous studies focused on optimizing the
energy efficiency in rail transportation and evaluation of the
alternative energy sources that would be critical for the AT
development. For example, Brenna et al. [73] indicated that
many European countries pursue the environmental
sustainability and energy efficiency goals. One of the
alternatives to improve the environmental sustainability of
rail transportation is the AT deployment. The study
developed a genetic algorithm to optimize the energy
consumption for autonomous subway trains. Along with the
energy consumption, the fitness function of the algorithm
minimized the total delay. The energy optimization was
achieved through the control of train movements (e.g.,
setting the train speed, determination of stop positions). The
proposed methodology was applied for a real-life subway
line in Milan (Italy) and demonstrated substantial energy
savings.
Mandara et al. [74] presented a prototype of the fully
automated metro train with improved safety features. The
improved safety level was achieved by introducing a
monitoring unit that could detect potential safety hazards on
rail tracks in front of the train. The Li-Fi technology was
suggested for communication purposes. The roof of the
proposed train system was assumed to have a set of solar
panels, so the renewable energy source could be partially
used instead of solely relying on the electric power. The
proposed train system also deployed a set of sensors to
estimate the weight of passengers inside the train and
ensure that the train capacity is not violated. It was
highlighted that the proposed driverless train system could
effectively prevent potential issues caused by human errors.
Richert [75] described the new generation trains that would
have the hydrogen fuel cells and storage batteries as power
sources. The East Japan Railway Company, Toyota Motor
Corporation, and Hitachi, Ltd. entered into collaboration to
develop the prototypes of such trains. The main objective of
the project is to improve the environmental sustainability of
rail transportation and its competitiveness as compared to
other modes.
As indicated earlier, innovative AI-based techniques are
critical for the future development and deployment of ATs.
A number of the collected studies deployed the AI-based
methods for addressing some of the issues that are
associated with the train design and operations.
Gschwandtner et al. [76] highlighted some of the main
operational differences that could affect object detection on
highway streets and rail tracks. The study proposed a
method of using lane detection techniques for ATs in order
to effectively detect the surrounding obstacles on rail
tracks. The developed algorithm combined various
techniques used in lane detection and strong geometric
constraints that are specifically applicable for rail tracks.
The imposed geometric constraints decreased the
processing cost and produced robust outputs. However, the
accuracy of the proposed method could be influenced by
custom rail properties. Weichselbaum et al. [77] proposed a
3-D vision-based obstacle detection system that could be
used by ATs in open terrain environments. A number of
modifications were applied in the system to improve its
performance on a high-speed stereo engine. Stereo
matching and slanted correlation masks were able to
substantially improve the obstacle detection rate (i.e., from
89.4% to 95.75%). The false positive detection rate
comprised 0.25% only. The developed system was
evaluated using the real-world test data and proved its
effectiveness.
Xie et al. [78] highlighted that the adoption of the
European Rail Traffic Management System (ERTMS) and
the AT deployment could be considered as two main
solutions to enhance the safety of rail operations and
effectively increase the existing capacity to ensure that the
demand is met. The study proposed a methodology
allowing the application of discrete event controllers for the
AT control system in rail transportation. The modeling was
performed using the Colored Petri Nets along with its
extensions, considering some important aspects (e.g., rail
operational requirements, collision-free systems). It was
found that the proposed methodology could be efficient in
emulating AT control systems and similar complex
systems. Talvitie et al. [79] addressed the positioning of
high-speed train within 5G new radio networks by means of
new radio synchronization signals. The proposed
positioning method took into consideration the time of
arrival along with the angle of departure, which were both
computed using new radio synchronization signals. The
Extended Kalman Filter was used to track the train position
based on the assumed train movement model and given
measurements. It was found that the time of arrival
measurements were able to provide better accuracy, when
comparing to the angle of departure measurements.
Moreover, the best accuracy was accomplished when both
measurements were taken into account. The proposed train
positioning method could be useful in a variety of rail
applications, including the AT operations.
Yin et al. [18] presented a state-of-the-art review on the
AI applications in high-speed rail transportation. The study

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
21
outlined a variety of areas where the AI-based techniques
could be used to improve rail transportation, including
timetable synchronization and optimization, knowledge-
based customer service, speed control and trajectory
control, intelligent equipment, and intelligent maintenance.
Many of the aforementioned AI applications would be
critical for the AT development as well. Furthermore, a
comprehensive framework was developed for the AI
applications in high-speed rail transportation, considering
operational efficiency and customer comfort criteria. Along
with the aforementioned efforts, many other studies applied
various AI-based methods, optimization, simulation, and
other operations research techniques to address different
decision problems and issues that are related to train
operations, including energy and power sources [80-96],
train speed control and trajectory control [97-106],
timetable synchronization and optimization [107-117], as
well as intelligent maintenance and prognostics [118-127].
E.
AUTONOMOUS TRAINS AND HIGHWAY-RAIL
GRADE CROSSINGS
A highway-rail grade crossing (HRGC), also known as ―a
level crossing‖, represents the location, where a highway
segment intersects with a railroad segment at the same
elevation. Each HRGC creates a conflicting point between
the arriving trains and vehicles. A significant number of
accidents involving trains and vehicles at HRGCs are
reported every year in many countries [128-131]. A number
of studies have been conducted to enhance safety at
HRGCs, which would be critical for the development and
deployment of the future ATs. Hsu and Jones [132]
analyzed the CV transmission range requirements at
HRGCs, aiming to improve the overall safety of
approaching vehicles and passing trains. The study assumed
that the approaching trains had the onboard communication
units that transmitted the information regarding the train
speed and location to the CVs. The time-to-collision and
stopping distance were used in the reliability-based risk
analysis. A set of Monte Carlo simulations were executed
to determine the risk probabilities that a CV would not stop
before the train arrival time or within the transmission
range. The results from the conducted analysis
demonstrated that the CV collision risk with a 600-meter
transmission range would be lower, when comparing to the
non-CVs, for the HRGCs with passive protection and 300-
meter sight distance to the train. Moreover, the CV collision
risk was found to be lower for the HRGCs with active
protection and a 300-meter transmission range. It was
concluded that a provision of 600-meter transmission range
would be a feasible alternative to enhance the safety levels
at passive and active HRGCs.
Zaouk and Ozdemir [133] underlined that the accidents
between vehicles and trains at HRGCs cause a significant
number of fatalities and injuries. The study mainly
concentrated on the following aspects in order to improve
safety at HRGCs: (1) interaction of drivers with crossings
and traffic patterns; (2) systematic collection of accident
data and its evaluation; (3) enforcement and regulations; (4)
public perception and education; (5) institutional issues;
and (6) technological development and its modernization. It
was highlighted that the DSRC technology could be
effectively integrated with HRGC systems to prevent
potential safety issues and improve traffic flows via
HRGCs. In particular, the status of an HRGC may be
directly broadcast to approaching CVs, so they could take
the appropriate actions (e.g., slow down or completely stop
because of an approaching train). One of the major
challenges with the development of CV and AV
technologies was found to be its limited testing in various
environments (e.g., HRGCs with passing ATs and
conventional trains) under various operational scenarios.
Effective testing of the new technologies at HRGCs is
critical to make sure that the approaching vehicles and
passing trains would be able to safely interact with each
other.
Voege et al. [134] indicated that a high level of safety is
viewed as one of the main operational goals of rail
transportation. The study highlighted that automation is
expected to eliminate human errors, which is critical to
ensure safety at HRGCs (as collisions between approaching
vehicles and passing trains often occur as a result of human
errors). It was underlined that the ATs with a full
automation level would be able to control the appropriate
speed levels at different segments of the rail line. New
generation AVs would also have the speed control features
to make sure the appropriate speed would be selected when
approaching HRGCs. Moreover, the introduction of new
maintenance concepts (e.g., predictive maintenance) would
allow reducing the probability of accidents due to
deteriorating conditions of certain railroad assets. U.S.
DOT [135] listed the main principles of automation,
including the following: (1) safety prioritization; (2)
adaptation of technologically neutral policies; (3)
elimination of outdated regulations; (4) consistent
regulatory and operational environment; (5) proactive
preparation for automation; and (6) protect the freedom for
citizens to drive (e.g., some users may be still willing to
have manually-driven vehicles). The study highlighted the
importance of effective interactions between AVs and
passing trains at HRGCs. A variety of stakeholders are
involved to design a safe system for AV-train interactions.
U.S. DOT [136] aimed to determine a set of primary
requirements for AVs to safely pass through HRGCs. The
study pointed out that the development of new generation
technologies is constrained by the uncertainties and
differences in perspectives between the relevant
stakeholders and industries (e.g., automated driver
assistance systems, Federal Highway Administration, and
railroad authorities). It was recommended that the
representative group of stakeholders should reach a
consensus on a sufficiently broad set of requirements for
AVs to safely pass through HRGCs. The CAV technology
could serve as an effective alternative for improving safety
of users at crossings. Bertini and Wang [137] and the U.S.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
22
Government Accountability Office (GAO) [138] also
pointed out that the CAV technology could improve
interactions between approaching vehicles and passing
trains at HRGCs. Neumeister et al. [139] discussed the Rail
Crossing Violation Warning (RCVW) system, which was
introduced to improve safety at HRGCs. Based on the
RCVW concept, the road-side units communicate to the
CVs regarding the approaching train. The RCVW systems
rely on different components, including computing
platforms, DSRC, GPS, and driver-vehicle interface. The
study highlighted that the initial RCVW prototype would be
steadily improved to meet the performance standards
through an extensive field testing.
FIGURE 17. Approaching vehicles and passing AT at an HRGC.
Virtanen et al. [140] studied the issues associated with
the AVs passing through HRGCs with and without any
protection. The use of AV perception sensors, vehicle-to-
everything messaging, and train-tracking solutions could
provide the required information for the AVs to safely pass
through HRGCs. It was underlined that the protected
HRGCs can be detected with the AV perception sensors.
Furthermore, the messages could be sent to a vehicle using
road-side units in the vicinity of protected HRGCs (see Fig.
17). On the other hand, the AV perception sensors alone
may not be sufficient to detect an approaching train at
unprotected HRGCs, especially under inclement weather
conditions. The study presented a train-tracking solution
that could detect train arrivals and produce safety messages
for AVs. Wang et al. [141] proposed a new active warning
system that is based on the readily available CV
technologies to improve safety at HRGCs. The system
relied on a wireless communication via the DSRC between
the onboard equipment and road-side units to activate the
warning messages due to approaching trains. In case the
highway users are at greater risk, the system would send
visual and auditory alerts along with displaying the
expected waiting time. The suggested warning system was
evaluated using a simulation model and field applications.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
23
The results from the conducted experiments indicated that
the proposed active warning system could effectively
decrease potential collisions between approaching vehicles
and passing trains.
Knapp et al. [142] highlighted that it is critical for CAVs
to effectively sense the surrounding environment.
Therefore, there should be consistent standards for
designing the roadway and rail infrastructure (e.g.,
pavement markings should have the appropriate contrast
and width, so they can be easily detected by CAVs when
approaching HRGCs). The Manual on Uniform Traffic
Control Devices (MUTCD) provides a comprehensive
guidance on the use of different traffic control device types
in the U.S. Furthermore, the MUTCD provides details
regarding the requirements for pavement markings and
signage at HRGCs. In some countries (e.g., Canada), there
are other types of manuals that are specific to HRGCs. Li
and Liu [143] developed a comprehensive intersection
management strategy for effective V2V and V2I
communication. It was assumed that AVs could check the
status of an intersection and identify traffic congestion by
means of V2I communication. A static conflict matrix was
used to determine the priority of passing a given
intersection. Without loss of generality, the proposed
methodology could be extended for V2I communication in
the vicinity of HRGCs, so the approaching AVs would have
the information regarding the status of a given HRGC.
Rosyidi et al. [144] presented a radar-based sensor
system, which could be used to improve safety at HRGCs
and prevent potential collisions between trains and vehicles.
The radar-based sensor was deployed to identify the
approaching train and send the information to the micro-
controller, which could initiate gate closing and opening
operations at a given HRGC. In case of the approaching
train, the proposed HRGC crossing system also activates
warning lights to alert the pedestrians.
U.S. DOT [145] examined the best ways of how various
engineering firms, transportation agencies, researchers, and
other infrastructure stakeholders could effectively
collaborate and prepare the existing passenger and freight
rail infrastructure for the CV and AV deployment. The
following aspects were explored: (1) information
requirements for the vehicles approaching HRGCs; (2) the
existing barriers for the data collection and exchange
between different entities (e.g., public highway agencies,
private freight railroad operators, public transit/commuter
railroad agencies); (3) CV and AV implementation
approaches that could potentially benefit highway and rail
systems; and (4) additional research efforts needed to
address the CV and AV architecture issues and challenges.
The study pointed out that the existing PTC system would
not be able to effectively provide the train location
information to the vehicles approaching HRGCs in the
U.S., as the PTC system had been installed only on less
than 50% of the national rail miles. It was indicated that the
train automation and train crew size reduction could
facilitate the deployment of CV and AV technologies in the
vicinity of HRGCs.
U.S. DOT [146] underlined that more than $8 million
were invested in the U.S. into the automated transit
research, including the first fully automated bus rapid
transit system. Furthermore, around $60 million were
allocated for the deployment of automated driving systems
across the country. It was also indicated that more efforts
have been made towards modernizing the regulatory
environment of automated driving systems. The main goals
of automation were categorized into three groups: (1)
protect users and communities; (2) promote effective
markets; and (3) facilitate coordinated efforts. The proposed
comprehensive plan focused not only on CAVs but also on
the modal interface points (e.g., HRGCs, marine ports).
IV. REVIEW SUMMARY
TABLE IV summarizes the findings that were revealed
from the reviewed studies, including the following
information: (1) author(s) and year; (2) study objective; and
(3) notes, important considerations, and study outcomes.
Furthermore, the main advantages and challenges from the
AT deployment, which were identified as a result of the
conducted state-of-the-practice and state-of-the art review,
are presented in the following sections.
A.
ADVANTAGES FROM THE DEPLOYMENT OF
AUTONOMOUS TRAINS
As a result of the conducted state-of-the-practice and state-
of-the-art review, the following main advantages from the
AT deployment have been identified:
 ATs are expected to improve the overall safety
throughout the transportation process [16, 62]. Human
errors are viewed as one of the main causes of
accidents in rail transportation networks. The ATs with
a full automation level will not require any onboard
personnel, which will practically eliminate any
possibilities of human errors and associated accidents.
Fully automated systems and computer-based programs
are expected to control the AT movements more
precisely as compared to human operators. New
generation sensors, platform supervision technologies,
detection systems, and intrusion prevention systems
will guarantee an acceptable safety level.
 The AT deployment will allow increasing the existing
capacity and utilization of rail lines [42, 45, 52, 70,
78]. The capacity increase can be achieved by reducing
the headway between consecutive ATs. Unlike
conventional trains, ATs rely on different technologies
that would allow substantially decreasing the headway
without affecting the safety level. Shorter headways are
expected to increase the service frequency of metro
stations, reduce the waiting time of passengers at metro
stations, and facilitate the alighting and boarding
process. Some experts underline that the capacity
increase is one of the main reasons for rail automation,
as many conventional rail lines operate near their
capacity limits [62].

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
24
TABLE IV
SUMMARY OF THE REVIEWED STUDIES
a/a
Author(s) and Year
Study Objective
Notes, Important Considerations, and Study Outcomes
Connected and Autonomous Vehicle Technologies and Their Applications for Autonomous Trains
1
Bagloee et al.
(2016) [30]
Analyze various issues and
opportunities regarding the
transportation policies and regulations
associated with the CAV technologies.
An essential part of the CAV technology was found to be its ability to
communicate with moving vehicles and infrastructure. This technology would be
important for the vehicles passing through highway-rail grade crossings and
interacting with the trains approaching the crossings.
2
Krasniqi and
Hajirizi (2016) [31]
Investigate the current CAV market
and technology trends from the
preliminary stage to fully driverless
vehicles.
The IoT technology was found to have a potential to completely alter the
automobile market, which could provide a major thrust to IoT in the autonomous
technology market. The study concluded that the DSRC and 5G technologies
would be the most suited for improving safety and vehicle communication with
trains and surrounding infrastructure.
3
Crains (2017) [32]
Present a perspective on the future of
autonomous technology in all modes
of transportation.
A number of challenges with the AT deployment were pointed out (e.g.,
introduction of the PTC system in the U.S. for freight railways only; disagreement
of railroad unions to adaptation of one-person train crews), which can slow down
rail automation if not addressed properly.
4
Bucsky (2018) [33]
Study the existing condition of freight
traffic and CAV applications in freight
transportation.
The study highlighted that the railroad industry would have more developments
towards the full automation level. However, such developments would require
significant investments. It was indicated that the automation would make the road
transport more preferential as compared to the other modes, which might lead to
some environmental problems.
5
GHSA (2018) [34]
Determine potential issues from the
deployment of autonomous driving
systems.
The following safety issues were identified: compliance with traffic laws, decisions
during emergency situations, recognition, reaction, operations after detection of
certain system failures, system security, and others. It was recommended that
partially automated driving systems should be controlled by a licensed human
operator.
6
Elliot et al. (2019)
[36]
Detailed review of recent
advancements in CAV technologies.
The study highlighted the importance of the DSRC and 5G technologies for
effective communication between CAVs and trains. Addressing security issues,
reliability issues, and conflicts in operations would be an important next step in the
CAV deployment.
7
CRS (2020) [27]
Investigate the issues related to the
deployment and testing of CAVs.
The interaction of CAVs with the existing infrastructure, ATs, and non-CAVs will
generate a huge amount of data (e.g., vehicle location and train location). Any
unauthorized access may lead to safety issues not only for the vehicle and for the
train but for others as well.
General Issues Associated with Autonomous Trains
8
Gebauer and Pree
(2008) [37]
Determine the conceptual, technical,
and legal challenges associated with
the AT deployment.
The concept of ―trainlets‖ was proposed to improve the attractiveness of ATs. The
entire AT could be divided into many smaller ―trainlets‖, which have the size of a
standard automobile. The studies discussed the autoBAHN project that was
launched to validate the technical feasibility of fully autonomous ―trainlets‖. It was
highlighted that the economic viability of ATs would depend on how the
automation concept would be implemented. In particular, the use of the existing
railroad infrastructure with the minimum required alterations would be favorable
for the development and deployment of ATs. Given the existing interest of the
relevant stakeholders, the autoBAHN project might be implemented in the near
future.
9
Gebauer et al.
(2012a) [38]
10
Gebauer et al.
(2012b) [39]
11
Dordal and Avila
(2016) [40]
Develop a software agent-based
system to coordinate trains.
The results of the study showed that the proposed software agent-based system
could reduce the total journey time by 22.5%. The fuel consumption was reduced
by more than 25%, which could further improve the environmental sustainability.
12
Hazan et al. (2016)
[41]
Study the impacts of AV deployment
on rail transportation.
As a result of conducted survey, it was found that about 50% of respondents (out of
5,500 consumers from ten different countries) would consider buying AVs. More
importantly, at least 40% of the current train passengers would shift to using AVs.
13
Powell et al. (2016)
[42]
Assess benefits and challenges of the
AT deployment for the Tyne and Wear
Metro (U.K.).
Low adhesion conditions and the presence of shared operations at the existing rail
lines were found to be the main obstacles. Moreover, additional infrastructure
upgrades would be necessary to transition towards full automation.
14
Wang et al. (2016)
[16]
Review the driverless train operations
in urban transit rail systems.
Along with potential advantages, different AT deployment challenges were
outlined. A systematic safety assessment framework is needed to standardize the
AT operations, reduce potential risks, and enhance the overall reliability.
15
Trentesaux et al.
(2018) [43]
Investigate advantages and
disadvantages from the AT
deployment.
The deployment of ATs would significantly change the railroads. However, the
success of ATs will depend on the participation of various stakeholders, such as
manufacturers, fleet operators, infrastructure developers, etc.
16
Wardrop (2019)
[44]
Identify the challenges associated with
the freight train deployment in remote
areas.
Mining in remote areas is generally expensive, as it is quite difficult to find the
required resources. Therefore, the automation of mining process and rail
transportation gained popularity among mining companies.
17
Antolini (2020)
[45]
Determine potential challenges of the
AT deployment by the SNCF railroad
company.
The main challenge of the AT deployment in Europe would be the introduction of
ATs on the railroad system, where multiple passenger and freight operators share
the same railroad tracks. The trains might have different physical characteristics.
18
Kemmeter (2020)
[46]
Investigate the AT deployment
challenges.
Certain CAV technologies may not be applicable for ATs. Train detection requires
installation of cables along the rail tracks as well as a lot of online equipment. Such
train detection equipment is generally costly and requires regular maintenance.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
25
TABLE IV
SUMMARY OF THE REVIEWED STUDIES (CONT’D)
a/a
Author(s) and Year
Study Objective
Notes, Important Considerations, and Study Outcomes
19
Muller (2020) [47]
Evaluate the organizational hindrances
from the deployment of freight ATs.
The study evaluated four conceptual impacts on innovation activities within
advanced markets. It was concluded that the strategic innovation policies should
address the challenges that are associated with the AT deployment.
20
Pattinson et al.
(2020) [48]
Discuss and analyze the legal issues
associated with AVs.
It is critical to develop the appropriate solutions to determine legal responsibilities
for vehicles and drivers in case of accidents. Similar legal issues have to be
addressed when deploying ATs and human-driven trains on the existing rail lines.
21
Wang et al. (2020)
[49]
Study the CAV safety issues.
Drivers often take control of AVs when they think it is necessary, which creates a
―disagreement‖. Such disagreements may further lead to accidents, which can
occur at the rail lines with the grades of automation GoA2 and GoA3 as well.
22
Othman (2021) [50]
Conduct a comprehensive review of
the AV public acceptance and
perception.
The COVID-19 pandemic made substantial impacts on public transit services
around the world. The future rail transit systems, including autonomous rail transit
systems, should have additional protective measures against the spread of diseases.
23
Karvonen et al.
(2011) [51]
Investigate challenges from the AT
deployment in metro systems.
Train drivers are viewed as a link between different actors involved in different
metro operations. It was concluded that a lack of understanding of the train driver
roles may lead to safety and quality of service issues throughout the AT
deployment.
24
Cohen et al. (2015)
[52]
Determine the effects of automation
on metro line operations.
The results indicated that the AT deployment could decrease the train crew size by
30-70% (which can be negatively viewed by the public). The findings showed that
the AT deployment could improve the metro efficiency and productivity.
25
Cassauwers (2020)
[53]
Study the employment issues due to
the AT deployment.
The study pointed out that the re-orientation of employees (e.g., transition to
customer service or to non-automated rail lines) would be a promising solution
rather than layoffs or salary cuts.
User Perception and Outlook for Autonomous Trains
26
Fraszczyk et al.
(2015) [54]
Study public perception of substantial
changes in rail transport associated
with ATs.
The survey involved 50 individuals from 10 different countries. The majority of
survey participants indicated no concerns associated with the AT maintenance
issues and the train design itself. However, staff communication and technical
failures were the main two issues that were raised.
27
Fraszczyk and
Mulley (2017) [55]
Investigate the user perception and
attitude towards ATs in Sydney
(Australia).
The survey involved 300 individuals from Sydney (Australia). The study results
indicated that many train users and non-users rated the presence of a train driver as
important. The majority of survey participants were concerned with the safety
aspects of the AT operations.
28
Pakusch and
Bossauer (2017)
[56]
Study the user acceptance to
technological innovations in
autonomous public transportation.
The survey involved 201 individuals from Germany. Many users were already
familiar with autonomous driving and were willing to use autonomous public
transportation in the following years. The existing policies should be modified to
allow the users accessing autonomous public transportation during test phases.
29
Henne et al. (2019)
[57]
Address the user perception challenges
towards AI-based applications for
autonomous systems.
A combination of methods for assessing the perception uncertainty and dynamic
dependability management was found to be a promising solution to effectively
address the challenge of unreliable perception of the AI-based applications for
autonomous systems.
Design and Technologies for Autonomous Trains
30
Bock and Bikker
(2000) [58]
Develop a new operational concept for
rail services.
The proposed concept assumed that the train wagons were no longer physically
coupled to each other. Each wagon had its own computer system along with
propulsion, which can be considered as an AT module prototype.
31
Haxthausen and
Peleska (2000) [59]
Propose a controlled distributed
system for railroads.
The proposed concept could be used for an automatic train control without the
presence of a train driver (i.e., AT applications in rail transportation).
32
Matsumoto et al.
(2001) [60]
Design a decentralized autonomous
train control (D-ATC) system.
The proposed D-ATC system had the following main features: (1) train detection
on rail tracks; (2) dissemination of the stop point information; (3) detection of a
train position; and (4) control of braking.
33
Matsumoto (2005)
[61]
Review various train control systems
that were deployed in Japan.
The newly designed information technologies allow trains detecting their own
positions and transmitting this information to other trains in the vicinity.
34
Castells et al.
(2011) [62]
Study the operational AT features at
metro lines.
The AT signaling and control systems allow shorter headways between consecutive
trains, which increases the rail network utilization. The AT deployment allows
better management of the supply and demand.
35
Marrone et al.
(2012) [63]
Study the verification process of
different AT applications.
The study indicated that the Combined Model-Based Analysis and Testing of
Embedded Systems (MBAT) project conducted in the European Union would help
improving the verification process.
36
Mohammed et al.
(2014) [64]
Develop a microcontroller-based
prototype for ATs.
The proposed AT prototype was able to perform a set of basic operations (e.g.,
travel via a pre-defined path between a set of stations, proper stopping at stations).
The proposed AT prototype could serve as a foundation for more sophisticated
control systems.
37
Balasubramanian
(2016) [65]
Investigate various IoT processors that
could be utilized to deploy ATs for a
long-distance travel.
A successful implementation of the IoT technologies is heavily dependent on the
availability of high-speed internet. The proposed concept of long-distance ATs is
expected to have a variety of advantages.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
26
TABLE IV
SUMMARY OF THE REVIEWED STUDIES (CONT’D)
a/a
Author(s) and Year
Study Objective
Notes, Important Considerations, and Study Outcomes
38
Fraga-Lamas et al.
(2017) [17]
Review the evolution of railroad
communication technologies.
The IIoT and Internet of Trains still face a large variety of issues (e.g.,
interoperability, standardization, scalability, and cybersecurity) that have to be
effectively addressed by researchers and practitioners in the following years. New
types of technology should be tuned for specific railroad environments to ensure
their adequate implementation.
39
Romano et al.
(2017) [66]
Study the existing AT metro systems
and changing demand.
The new generation ATs should have the appropriate types of technology that
could achieve operational cost optimization, energy reduction targets, minimization
of risks associated with human errors, and high system performance.
40
Bharathi et al.
(2018) [67]
Design a technology for automatic
train control and operations.
The train was assumed to have a controller enabling the train automatically move
from one metro station to another (i.e., an AT prototype). Upon arrival at a metro
station, the train could stop automatically by applying the appropriate sensors.
Such an automatic train system could be effective as it eliminates potential human
errors.
41
Lagay and Abdell
et al. (2018) [68]
Discuss various features of the AT
deployment project in France.
A variety of AT technologies and systems were described. In order to effectively
deploy ATs, the SNCF railroad company planned to work with different industrial
and service sectors, as well as low-cost solutions and high-performance solutions.
42
Arup (2019) [69]
Review new trends in rail
transportation.
The AT technologies allow optimizing the running time, increasing the average
speed, and operations in a close proximity to other trains. The AT deployment also
allows effectively addressing the growing demand for rail transportation.
43
Kimiagar (2019)
[22]
Review the emerging trends in the
technology used for rail transit.
The study presented different train control systems that are responsible for various
functions (e.g., train timetable development, ride comfort, safety aspects, and
interlocking). The steps required for the transition from conventional trains to ATs
were pointed out as well.
44
Pickering et al.
(2020) [70]
Design a simulation model to
investigate the AT deployment.
The results from the conducted simulation analysis demonstrated that the AT
deployment could substantially increase the rail network capacity, as consecutive
trains could be operated at a smaller distance.
45
Kunifuji et al.
(2009) [71]
Design an alternative railroad
signaling system.
A new railroad signaling system was based on the network and autonomous
decentralized technology. The developed system would improve not only the
flexibility of signaling but the overall safety and reliability as well.
46
Harb et al. (2020)
[72]
Develop a dataset for railroad traffic
light recognition and detection.
The developed dataset can be used for automatic classification of railroad signaling
panels by state and by type. Furthermore, the developed dataset could assist to
facilitate the AT deployment.
47
Brenna et al. (2016)
[73]
Optimize the energy consumption for
autonomous subway trains.
The study developed a genetic algorithm to optimize the energy consumption for
autonomous subway trains. The energy optimization was achieved through the
control of train movements (e.g., setting the train speed, determination of stop
positions).
48
Mandara et al.
(2019) [74]
Suggest a prototype of the fully
automated metro train with improved
safety features.
The improved safety level was achieved by introducing a monitoring unit that
could detect potential safety hazards on rail tracks in front of the train. The roof of
the proposed train system was assumed to have a set of solar panels.
49
Richert (2020) [75]
Present new generation trains with
alternative power sources.
The new generation trains would have the hydrogen fuel cells and storage batteries
as power sources, which is expected to improve the environmental sustainability of
rail transportation and its competitiveness as compared to other modes.
50
Gschwandtner et al.
(2010) [76]
Develop a method of using lane
detection techniques for ATs.
The developed algorithm combined various techniques used in lane detection and
strong geometric constraints that are specifically applicable for rail tracks. The
accuracy of the proposed method could be influenced by custom rail properties.
51
Weichselbaum et
al. (2013) [77]
Propose a 3-D vision-based obstacle
detection system for ATs.
Stereo matching and slanted correlation masks were able to substantially improve
the obstacle detection rate (i.e., from 89.4% to 95.75%). The false positive
detection rate comprised 0.25% only.
52
Xie et al. (2017)
[78]
Develop discrete event controllers for
the AT control system.
The modeling was performed using the Colored Petri Nets along with its
extensions, considering some important aspects (e.g., rail operational requirements,
collision-free systems).
53
Talvitie et al.
(2018) [79]
Propose a train positioning method.
It was found that the time of arrival measurements were able to provide better
accuracy, when comparing to the angle of departure measurements. Moreover, the
best accuracy was accomplished when both measurements were taken into account.
54
Yin et al. (2020)
[18]
Review the AI applications in high-
speed rail transportation.
The study outlined a variety of areas where the AI-based techniques could be used
to improve rail transportation. A comprehensive framework was developed for the
AI applications in high-speed rail transportation, considering different criteria.
Autonomous Trains and Highway-Rail Grade Crossings
55
Hsu and Jones
(2017) [132]
Analyze the CV transmission range
requirements at HRGCs.
It was concluded that a provision of 600-meter transmission range would be a
feasible alternative to enhance the safety levels at passive and active HRGCs.
56
Zaouk and Ozdemir
(2017) [133]
Investigate various CAV-based
alternatives to improve safety at
HRGCs.
One of the major challenges with the development of CV and AV technologies was
found to be its limited testing in various environments (e.g., HRGCs with passing
ATs and conventional trains) under various operational scenarios.
57
Voege et al. (2017)
[134]
Investigate different aspects and
effects of automation.
Automation is expected to eliminate human errors, which is critical to ensure safety
at HRGCs (as collisions between approaching vehicles and passing trains often
occur as a result of human errors).

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
27
TABLE IV
SUMMARY OF THE REVIEWED STUDIES (CONT’D)
a/a
Author(s) and Year
Study Objective
Notes, Important Considerations, and Study Outcomes
58
U.S. DOT (2018a)
[135]
Discuss various features of automation
and how to prepare for the future.
The study highlighted the importance of effective interactions between AVs and
passing trains at HRGCs. A variety of stakeholders are involved to design a safe
system for AV-train interactions.
59
U.S. DOT (2018b)
[136]
Determine a set of primary
requirements for AVs to safely pass
through HRGCs.
The development of new generation technologies is constrained by the
uncertainties and differences in perspectives between the relevant stakeholders and
industries (e.g., automated driver assistance systems, Federal Highway
Administration, and railroad authorities).
60
Bertini and Wang
(2016) [137]
Prepare a roadmap for the CAV
applications in Oregon (U.S.).
The CAV technology could improve interactions between approaching vehicles
and passing trains at HRGCs.
61
GAO (2018) [138]
Analyze the safety challenges at
HRGCs.
The CAV technology could improve interactions between approaching vehicles
and passing trains at HRGCs.
62
Neumeister et al.
(2019) [139]
Design a Rail Crossing Violation
Warning (RCVW) system.
Based on the RCVW concept, the road-side units communicate to the CVs
regarding the approaching train. The RCVW systems rely on different components,
including computing platforms, DSRC, GPS, and driver-vehicle interface.
63
Virtanen et al.
(2019) [140]
Study potential issues associated with
the AVs passing through HRGCs.
Protected HRGCs can be detected with the AV perception sensors. The AV
perception sensors alone may not be sufficient to detect an approaching train at
unprotected HRGCs, especially under inclement weather conditions.
64
Wang et al. (2019)
[141]
Proposed a new active CV-based
warning system.
The system relied on a wireless communication via the DSRC between the onboard
equipment and road-side units to activate the warning messages due to approaching
trains. The system can send visual and auditory alerts along with displaying the
expected waiting time.
65
Knapp et al. (2020)
[142]
Review the codes and standards for
the future AV deployment.
The MUTCD provides a comprehensive guidance on the use of different traffic
control device types in the U.S. Furthermore, the MUTCD provides details
regarding the requirements for pavement markings and signage at HRGCs.
66
Li and Liu (2020)
[143]
Propose a strategy for the intersection
management, considering the presence
of AVs.
It was assumed that AVs could check the status of an intersection and identify
traffic congestion by means of V2I communication. The proposed methodology
could be extended for V2I communication in the vicinity of HRGCs
67
Rosyidi et al.
(2020) [144]
Improve safety at HRGCs.
Presented a radar-based sensor system, which could be used to prevent potential
collisions between trains and vehicles. The radar-based sensor was deployed to
identify the approaching train and send the information to the micro-controller.
68
U.S. DOT (2020a)
[145]
Investigate the best way to prepare for
the CV and AV deployment.
The existing PTC system would not be able to effectively provide the train location
information to the vehicles approaching HRGCs in the U.S., as the PTC system had
been installed only on less than 50% of the national rail miles.
69
U.S. DOT (2021)
[146]
Develop a comprehensive plan for the
AV deployment.
More than $8 million were invested in the U.S. into the automated transit research.
The proposed comprehensive plan focused not only on CAVs but also for the
modal interface points (e.g., HRGCs, marine ports).
 Although automation requires substantial initial
investments, ATs have lower operational costs as
compared to conventional trains [16, 62]. Operational
cost savings stem for the train crew size reduction (i.e.,
ATs with a full automation level will not require any
onboard staff) and the associated management,
training, and labor costs. For example, some of the
automated metro lines in Paris (France) were able to
reduce the operational costs by approximately 30% due
to the AT deployment [147]. The operational costs
savings due to the AT deployment will depend on
physical and operational characteristics of a rail line.
 The AT deployment is expected to improve the overall
service reliability [16, 52, 71], which is considered as
an important aspect for passenger and freight rail
transportation. An increase in the level of automation
allows reducing potential disruptions in train operations
due to human errors. The future ATs are anticipated to
detect disruptions in the planned operations and
respond promptly to ensure the comfort and safety of
passengers. Furthermore, the ability to adjust the AT
running times and more precise dwell times at metro
stations or freight terminals would make ATs more
reliable as compared to conventional trains. The service
reliability is also enhanced by increasing the service
availability, as metro stations can be visited more
frequently due to reduced train headways.
 Automation will allow improving service flexibility
and fleet management [16, 62]. More ATs can be
deployed during peak hours to make sure that the
demand is effectively met and reduce the waiting time
of passengers at metro stations. On the other hand,
unused ATs could be easily removed from the network
during off-peak hours. The AT fleet management and
removal of unused ATs from rail lines can be
effectively performed, as ATs are not human-driven.
An effective fleet management will also facilitate
timely maintenance (e.g., unused ATs can undergo
scheduled maintenance procedures).
 The introduction of ATs in the existing rail networks
will improve the energy efficiency [17, 43, 73]. ATs
are expected to consume less energy as compared to
conventional trains, since acceleration, traction, and
braking procedures will be optimized for ATs by
means of computer-based and AI technologies.
Similarly, the energy consumption by the AT air-
conditioning system can be optimized as well by
introducing the appropriate sensors that can control the

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
28
amount of needed fresh air [148]. Furthermore, the
LED lighting systems may be introduced within ATs to
reduce the energy consumption for lighting [149]. The
new generation ATs will also rely on renewable energy
sources, which will allow reducing the electric energy
consumption [16, 74]. Since ATs can visit metro
stations more frequently due to shorter headways, the
actual AT length can be decreased. The deployment of
shorter ATs can also decrease the energy consumption.
B.
CHALLENGES FROM THE DEPLOYMENT OF
AUTONOMOUS TRAINS
As a result of the conducted state-of-the-practice and state-
of-the-art review, many different challenges from the AT
deployment have been identified and classified into the
following categories: (1) design challenges; (2) operational
challenges; (3) technology-related challenges; and (4)
human aspect-related challenges. The identified challenges
are discussed in the following sections.
1) DESIGN CHALLENGES
 Achieving a full automation level for the existing
passenger and freight rail lines will require substantial
initial investments [16, 45, 46, 52]. ATs rely on
modern technologies, many of which are quite
expensive to obtain. The cost of installing the PTC and
DSRC technologies for effective communication is also
fairly high. The approximate cost of the PTC
technology can go up to $35K per locomotive, while
the approximate cost of the DSRC road-side unit can
go up to $40K [150]. Some experts indicate that the
cost of automation might be one of the main barriers
for the future AT development and deployment [33].
 Although the automation of rail transportation is
viewed as less complicated to implement as compared
to road transportation, the AT risk management issues
do exist. In particular, the AT operations in emergency
situations have to be investigated more in depth. The
AT braking distance due to emergency situations has to
be optimized (i.e., generally larger weight of trains
requires longer stopping distance) to prevent a high
impact due to collisions and unforeseeable activities on
tracks, such as intrusion by animals or trespassing [43,
53]. Adequate design improvements have to be made to
ensure a proper response of ATs in emergency
situations.
 Along with the automation of trains, the relevant
stakeholders should increase the level of automation
for maintenance procedures. Some of the critical
maintenance procedures that could be fully automated
include ballast replacement, ballast tamping, as well as
track relaying [17]. Fully automated maintenance
procedures are expected to ensure the adequate
operational conditions of the rail infrastructure and
avoid potential delays of ATs along the rail lines.
 One of the advantages from the AT deployment is an
increase in the capacity and rail line utilization due to
shorter headways between consecutive ATs. Due to
shorter headways, the rail terminal designs have to be
upgraded to ensure that the train turnaround
requirements are met [16]. The future AT terminals
should have specific areas designated for train storage
(when they are not being deployed), coupling and
decoupling of train units, and maintenance procedures.
 The appropriate stakeholders should focus on the
development of consistent codes and standards for
designing the roadway and rail infrastructure (e.g.,
pavement markings should have the appropriate
contrast and width, so they can be easily detected by
CAVs when approaching HRGCs) [142]. An improved
design of the roadway and rail infrastructure would
allow CAVs and ATs to effectively sense the
surrounding environment.
 Another significant challenge that substantially slows
down the AT development and deployment is
associated with legal issues [37-39]. The existing laws
and regulations put a lot of emphasis on safety and
security and require extensive testing of the AT
technologies before they could be implemented in
practice. Railroad companies, researchers, government
representatives, and other relevant stakeholders should
collaboratively develop more effective policies that
could facilitate the AT development and deployment
considering the perspectives of future users and
without affecting the safety level.
2) OPERATIONAL CHALLENGES
 ATs are being continuously developed and upgraded in
many countries. However, it will not be possible to
fully automate all passenger and freight rail lines at the
same time (i.e., manually-driven trains will have to
share rail lines with ATs, which may create some
operational challenges) [42, 45]. New policies and
operational strategies should be developed to improve
coordination of trains at shared rail lines, so that
manually-driven trains can co-exist with ATs.
 ATs rely on a wide range of different systems, and the
interoperability between these systems should be
steadily enhanced. Otherwise, the future ATs will not
be able to reach their full potential in the operational
effectiveness. One of the approaches the can be used to
improve the interoperability of AT systems is the
implementation of semantic data models [17]. The
semantic data models allow an effective integration of
the data generated by various systems.
 The future research should investigate different
coordination mechanisms between various
autonomous systems (e.g., how to minimize the total
waiting time of autonomous buses that expect the
arrival of certain passengers that are travelling on an
AT). Without effective coordination mechanisms the
users of autonomous systems may not be able to
experience all the benefits of automation. Furthermore,
a lack of effective coordination mechanisms may lead
to certain operational deficiencies (e.g., excessive idle
time of autonomous buses, ATs, and AVs).

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
29
 Operations scheduling problems generally have high
computational complexity, and efficient solution
algorithms are required to solve these problems [17,
151-153]. To ensure a successful AT deployment,
many different planning and scheduling problems
have to be solved, including train line balancing,
timetable development and optimization, real-time train
rescheduling, train reordering, and track maintenance
[154-159]. The future studies should concentrate on the
development of efficient solution algorithms for these
decision problems.
 The share of freight rail transport is fairly modest,
when comparing to other transportation modes [17].
Automation may not be able to fully address this issue,
and additional actions will be required from the
relevant stakeholders. One alternative that can be
considered is to improve the service quality and ensure
timely deliveries to the end customers. Moreover, the
introduction of cost-competitive services and
interoperability improvements are also viewed as
viable options.
3) TECHNOLOGY-RELATED CHALLENGES
 There is a continuous evolution process in rail
signaling systems, where communication-based train
control systems (e.g., radio technology, microwave
technology) are being used instead of track-circuit
signaling [16]. Additional steps and procedures have to
be undertaken by the relevant stakeholders to ensure a
smooth transition from one rail signaling system to
another.
 Despite the fact that the existing ATs rely on a variety
of AI-based technologies [18, 22, 25], more research
still has to be done in order to improve the
informational, decisional, and learning processes.
These processes have to be performed considering a
variety of physical and operational attributes, including
the train speed and location, speed and location of other
trains on a given rail line, presence of objects on tracks,
status of railroad signals, and others. Effective
informational, decisional, and learning processes will
improve the reliability of AT operations.
 More efforts should be geared towards improving
cybersecurity of ATs. Cyber-attacks may substantially
disrupt the AT operations [17], as ATs heavily rely on
the computer-based AI technologies. Such disruptions
may negatively affect the comfort of passengers or
even lead to safety issues. Therefore, robust
cybersecurity measures should be developed in the
future to prevent unauthorized access in the AT
computer systems. Moreover, the future ATs should be
programmed accordingly; so the required software
updates are regularly performed, and the AT computers
remain resilient to the new and already-known cyber-
threats.
 The future research should focus more on enhancing
the existing wireless sensor networks, especially in the
railroad environments. There exist several issues
associated with wireless sensor networks that have to
be addressed, including communication reliability, fast
transmission rates, measurement of vibrations,
management of high-volume data, energy harvesting,
energy efficiency, and data fusion [17].
 Throughout the AT operations, there is always a risk of
intrusion of people or objects on rail tracks [16]. The
existing communication and surveillance technologies
have to be enhanced to make sure that unexpected
objects on rail tracks will be detected in a timely
manner, so ATs could respond properly (e.g., make an
emergency stop due to trespassing). This will also help
improving safety and security of passengers.
 The COVID-19 pandemic made substantial impacts on
public transit services around the world and caused a
significant reduction in ridership [50]. The future rail
transit systems, including autonomous rail transit
systems, should have additional protective measures
against the spread of airborne diseases (e.g., advanced
air circulation, ultraviolet light disinfection). Such
measures will help preventing the impacts of airborne
diseases on rail transit system operations and improve
safety of passengers.
4) HUMAN ASPECT-RELATED CHALLENGES
 Train drivers perform many different functions (i.e.,
anticipation, observation, interpretation, and reaction to
events). Moreover, train drivers are viewed as a link
between different actors involved in various rail
operations [51]. There still exists a major challenge in
understanding all the roles performed by train drivers
and how these roles can be performed by ATs. The
safety level and quality of service may drastically
decline if ATs are not able to perform the main train
driver roles.
 User perception remains one of the AT deployment
barriers. Based on the conducted review, many users
still have concerns regarding the AT performance in
emergency situations without the onboard staff [54,
55]. Additional educational programs should be
developed and administered, so the future users will be
aware of how ATs operate not only under normal but
also under disruptive conditions and emergency
situations as well. Furthermore, the existing policies
should be modified to allow the users accessing
autonomous public transportation systems even during
the test phases, so they could become more familiar
with new technologies and have positive experience in
the future [56].
 The existing communication systems within ATs must
be improved. The ATs used on metro rail lines should
not only be able to communicate with the surrounding
infrastructure and other ATs but with the onboard
passengers as well. Generally, train drivers inform
passengers in case of changes in the planned schedule
or emergency situations [16]. Since ATs won’t have
any onboard staff, the pertinent information should be
transmitted by the operational center directly into ATs

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3091550, IEEE Access
Singh et al.: Deployment of Autonomous Trains in Rail Transportation: Current Trends and Existing Challenges
30
in the form of the voice messages and/or text messages
and/or videos, so the passengers will have the pertinent
information throughout their journey.
 One of the main concerns with the AT deployment in
many countries is associated with the employment
issues. Based on the existing projections, the AT
deployment can reduce the train crew size by 30-70%
[52] and result in layoffs [32, 53]. Layoffs may further
cause a large number of strikes by railroad unions. The
employment issues due to the AT deployment have to
be addressed by the appropriate stakeholders in the
nearest future, as they may substantially slow down the
AT development and deployment. The re-orientation of
employees (e.g., transition to customer service or to
non-automated rail lines) would be a promising
solution rather than layoffs or salary cuts [53].
V. CONCLUDING REMARKS AND FUTURE RESEARCH
NEEDS
The statistics shows that the rail network has seen a
tremendous growth globally in the last two decades. The
growth has been observed in terms of the number of
passengers traveling and in terms of freight movements as
well. The total length of rail miles increased from 1,099,685
km in 2004 to 1,142,890 km in 2018. The growth in rail
passenger and freight traffic along with a continuous rail
network expansion requires railroad companies making
improvements in the existing operations to maintain
profitability and effectively tackle the demand for rail
transportation. Automation is expected to effectively
address the growing demand for passenger and freight
transportation, safety issues, additional costs, environmental
problems, human errors, and increasing congestion. The
growth of autonomous vehicles using the state-of-the-art
connected vehicle technologies has paved the way for the
development of passenger and freight autonomous trains
(ATs), also known as driverless trains. <