Introducing Scaled Model Development to on-Sight Automatic Train

Operation

Tobias Hofmeier

1,2

and Martin Cichon

Institute of Vehicle Technology, Nuremberg Institute of Technology, Keßlerplatz 12, Nuremberg, Germany

Institute of Vehicle System Technology, Karlsruhe Institute of Technology, Rintheimer Querallee 2, Karlsruhe, Germany

Keywords: Automatic Train Operation, ATO, Scaled Model, Scaled Development, Scaled Technology, Scaled Testing.

Abstract: Rail systems are often not an economic option in terms of flexibility and cost in intermodal competition. To

address this issue, there is a push towards implementing automation and digitalization components. In recent

years, there has been a strong focus on on-sight automated train operation systems. As the applications move

beyond protected areas such as metro systems, to complex on-sight driving scenarios, the demands on system

development, verification, and validation increase. Methods from the automotive industry are well known for

overcoming these challenges with virtual development and final field testing. Fundamentally different

operating conditions prevent sufficient field testing, as rail infrastructure and vehicles are difficult to procure

for development and testing purposes. In science and research, scaled models are being promoted for similar

problems. These models allow for simulations to be verified and favorable estimates to be made. This paper

demonstrates the possibility of using scaled model methods for the development of on-sight automatic train

operation (ATO) functions. A demonstrator of a highly automated shunting locomotive is being built as a

scaled model and equipped with sensors for environment detection and localization as well as communication

interfaces. The feasibility of ATO functions in the scaled model is demonstrated using defined use cases.

1 INTRODUCTION

The climate policy framework for the coming decades

requires a shift in traffic volumes from road transport

to more environmentally friendly alternatives, with

rail being one of the preferred solutions (Kaack et al.,

2018).

In order to create a competitive rail-based

alternative to road transport, it is necessary to analyze

the existing disadvantages of rail transport in

passenger and freight transport.

According to (Trepáčová et al., 2020) the

common disadvantages of passenger rail transport are

crowding; organization of the rail system; financial

cost; low number of trains; delays; train route,

boarding and exiting the train as well as luggage

transportation and safety issues.

In rail freight transport, a fundamental distinction

must be made between full train and single

wagonload transport. In full train transport, large

quantities of general cargo or bulk goods are

predominantly transported with the same origin and

destination, whereas in single wagonload transport,

trains with different load types, origins and

destinations of each wagon are transported. The more

flexible single wagon load transport competes with

road transport. Disadvantages against road transport

are less flexibility on the last mile, poor predictability

and high costs. The costs for the shunting process

account for a significant proportion of the total costs

of single wagonload transport, without any direct

value added (Guglielminetti et al., 2015).

Consequently, shunting processes offer potential

for optimizing rail freight transport.

Automation and digitalization are seen as driving

factors for increasing predictability, which has a

positive impact on both passenger and freight

transport and secondarily increases economic

efficiency and performance of railway systems

(Pourian, 2023), (European Union, 2019).

Based on this expectation, several automation

projects in passenger and freight transport have

already been successfully implemented. As

automated railway systems become more widespread,

these systems also need to be developed for use in

critical environments. Current methods do not

guarantee sufficient testing and development

possibilities to reliably bring systems to market

378

Hofmeier, T. and Cichon, M.

Introducing Scaled Model Development to on-Sight Automatic Train Operation.

DOI: 10.5220/0012691500003702

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 10th International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2024), pages 378-385

ISBN: 978-989-758-703-0; ISSN: 2184-495X

maturity. The following paper presents the potential

of scaled models for the problems involved in the

development of on-sight automatic train operation

(ATO) functions.

2 AUTOMATIC TRAIN

OPERATION

As already mentioned, several automation projects

are already successfully in operation. The grade of

automation (GoA) is defined in (IEC, 2006) and

describes the ascending takeover of tasks by an

automated system, ranging from on-sight driving

operation GoA0 to unattended operation GoA4.

To examine the influence on systems currently under

development, the operating conditions of some

example systems are described below.

The first GoA4 system was already in use in Kobe,

Japan in 1981 (Powell et al., 2016) and represents a

sealed off system. Subsequently, the underground

railways in European cities became increasingly

automated. Another milestone was the launch of the

U2 and U3 underground lines in Nuremberg, which

was the first to ensure mixed operation with

automated and non-automated lines since 2008

(Zasiadko, 2019). The closed and mostly intersection-

free system, with few clearly defined interfaces to the

outside world, means that subways can be automated

simply by monitoring the infrastructure at critical

locations.

A further step in the automation of rail vehicles is

the S-Bahn in Hamburg, where ATO over European

Train Control System (ETCS) is demonstrated.

Passenger operation is carried out with GoA2

(automated operation with driver monitoring) and

depot operation is carried out according to GoA4

(fully automated) (DB Systemtechnik, 2021).

Rio Tinto also operates highly automated rail vehicles

in Australia with its auto haul project. The automation

is realized via a radio remote control as a fallback

level. Safe operation is realized through the extra

protection of critical points such as level crossings

and inhabited areas. Collisions with wild or grazing

animals pose the greatest risk on most of the route.

The train detects impacts on the locomotive and

provides the supervisor with data to evaluate the

impact. The supervisor can decide remotely whether

the journey must be interrupted or can be continued

(Rio Tinto, 2017).

The systems listed here are located in more or less

protected environments and are operated fully

automatically if human or economic damage can be

ruled out. For the further spread of automated rail

systems, solutions must be developed that also work

in critical areas of application. This process is

explained below.

2.1 Developing ATO Systems

The examples show that the automation of railroad

systems is used in a wide range of applications. In

terms of railroad technology, the areas of application

can be divided into main line and on-sight operation.

In order to promote their automation, it is necessary

to work out the differences.

On the mainline trains are mainly operated using

train control systems featuring signals. Due to their

mechanical properties, the trains have very long

braking distances, meaning that the stopping distance

cannot be monitored by the vehicle. The signals are

used to divide the line into individual blocks to ensure

that trains stop in good time. The signal either blocks

the following block or releases it and transmits the

permitted speed. In manual mode, the driver reacts to

the signal position. this process must be taken over by

the system in automated mode (Pachl, 2020).

In Europe, ETCS is being introduced to automate

main lines. From equipment level 2, it transmits the

permitted speeds by radio, receives position

information and monitors the information generated.

At ETCS level 3, it will be possible to operate without

"fixed block" signals in order to increase line

utilization by means of dynamic virtual blocks.

(Schnieder, 2021).

ECTS already provides a functional automation

basis for the mainline. However, significant

investment in the infrastructure is still required for the

rollout across the entire rail system (European

Commission, 2023).

The 'driving-on-sight' includes more complex

operations such as depot trains, shunting,

approaching stops or mixed tram and road traffic. Due

to the significantly lower speeds, it is possible to

monitor the stopping distance from the vehicle.

When the vehicle is automated, it is therefore

necessary to replace the driver's visual, (acoustic and

sensory) perception with an automated system. This

step is accomplished by merging several sensor data.

Depending on the application, color cameras, infrared

cameras, LiDAR, radar or ultrasonic sensors are used

to replace the driver's visual capabilities (Tagiew et

al., 2022). In addition, it is necessary to locate the

locomotive in the track plan with high accuracy in

order to distinguish between obstacles in the

clearance gauge or objects by merging with the data

from the environment detection system. Since the

Introducing Scaled Model Development to on-Sight Automatic Train Operation

379

installed balises do not provide the necessary

accuracy, positioning systems consisting of

GPS/GNSS, IMU and odometry sensors are used.

The basic problems and the sensor setup used are

similar to autonomous driving in the automotive

sector. Therefore, it is advisable to examine closely

the development and testing methods used in the

automotive industry.

With the increasing complexity, not only the

development but also the adequate testing of

automated systems is becoming more and more

important. In the automotive industry, the scenario-

based test approach developed in (Pegasus, 2019) is a

common practice to support the development process

and bring the automated system to market.

In the early stages of the project, systems are

developed and tested in simulation environments; as

development progresses, the system is tested in the

real environment.

This approach seems appropriate for rail

technology. However, due to the fundamental

differences between road and rail vehicles in terms of

masses, friction coefficients, track characteristics and

operating environments, it is necessary to adapt the

approach for use in rail technology.

A correspondingly derived method for railroad

technology is described in (Greiner-Fuchs et al.,

2022). Due to the lack of general measurement data

with ATO reference in railway technology, the

methodology is driven by a knowledge database

derived from operational regulations, observations in

the field and self-generated data. Similar to the

automotive industry, the methodology propagates the

performance of simulative tests and final verification

and validation in the field.

In (Schäfer et al., 2023), the application of a

virtual environment for the development of highly-

automated rail vehicles is demonstrated. At the level

of research questions, it is therefore feasible to use

virtual environments to pre-develop systems. To date,

these simulations have not been sufficiently verified

and validated, so there is no commercially available

platform for on-sight ATO functions. The

development of field tests using the above-mentioned

method is also still under development and has not yet

been sufficiently validated.

It should therefore be noted that the fundamental

problems of highly automated rail vehicles are known

and the first solutions derived from the automotive

industry are available. The sensor concepts and

system topologies have been concretized, but

unfortunately the test and development standards do

not yet correspond to the desired target image. The

problems still to be solved are discussed below.

2.2 Challenges for Development and

Testing of on-sight ATO

Functions

Based on the research group's experience in the field

of automated on-sight rail systems, some remaining

problems for simulative and field testing can be

identified.

On the simulation side, there is the problem of the

dynamic development of new or other sensor

technologies for which adequate sensor models must

first be developed. Due to a long period of low

demand for simulation environments for rail vehicles,

the number of commercial participants is still very

low, which means that it is not always possible to

adapt the simulation environments in step with sensor

development. State-of-the-art development must

therefore often be carried out with non-validated

sensor models. With sufficient prior knowledge, the

basic sensor simulations can be considered reliable,

although they may not always provide a solid

foundation.

Railroad test sites are not public areas.

Consequently, clearly defined operational and

regulatory conditions apply on the available test sites.

This means that an operational and technical test

manager must be present in addition to the train driver

for the actual test and development personnel. The

high utilization of railway infrastructure also makes it

difficult to provide temporary routes for test and

development purposes without retrospective impact.

Maintenance backlogs for locomotives coupled with

the general availability of railway vehicles also make

field operations more difficult.

Prices for the use of a shunting area e.g. in

Germany are listed in (RLC Wustermark, 2023). A

locomotive including a shunting driver and diesel are

required for testing. This process can be considered a

shunting service for rough calculation purposes.

Additionally, costs are incurred for infrastructure,

such as track utilization. It is important to note that

the locomotive may lose its license when upgraded,

so costs for upgrading, disassembly, and certification

must be factored into the test. Accordingly, it is

advisable to plan test and development runs in the

field for extended periods to reduce the expenses and

labour involved in conversion and transportation.

Taking into account the daily rates for operational and

technical test managers, as well as the test and

development staff, the cost of a single test day can

amount to several thousand euros.

The high cost and inadequate availability of test

facilities and personnel ensure a high degree of

transferability of simulation results to the field.

VEHITS 2024 - 10th International Conference on Vehicle Technology and Intelligent Transport Systems

380

However, this is often not the case due to the

aforementioned non-validated simulation data, non-

simulated interfaces and unforeseen hardware effects.

In order to optimize the test time in the field, the

leap in technology maturity between simulation and

the real environment must be reduced. One possible

solution appears to be the use of scaled models, this

option is presented and discussed below.

Interface technologies such as WiFi, Ethernet or

radio and protocols such as TCP/IP, UDP, I²C, etc.

can still be used in their original form even at very

large downscaling due to their small size. This allows

interface problems, faulty data, bit shifts or similar

connectivity issues to be resolved realistically in the

model.

3 SCALED MODELS

Scaled models are commonly used in the literature

when real test objects are expensive or difficult to

obtain. They are also utilized to verify simulation

models. The reasons described are similar to the

problems observed during on-sight ATO-function

development and testing. The following section

shows how scale models are adapted in technology

and science to identify conditions and limits.

3.1 Scaled Models in Science

A large number of scaled models are known in science

and in engineering applications. These are presented

below as examples and analyzed with regard to

synergies with the problems discussed in the paper.

The investigation of air flows in wind tunnels is

one of the best-known applications of scaled models.

Large objects such as airplanes (Aerospace

Engineering, 2022) or buildings (Geurts & Van

Bentum, 2007) cannot be tested economically in their

original size. Consequently, the wind tunnel (test site)

and test object are scaled until a satisfactory

relationship between cost and gain in knowledge is

achieved. The measurement results are scaled to the

real application using physical correlations, such as

the Reynolds number. The scaled models are also

applied to verify simulation approaches.

Furthermore, scaled models are utilized for the

design of buildings for special cases such as

earthquakes. Various approaches are explored to

determine the scaling factors by which the results and

effects can be transferred to reality (Atar, 2022).

The research and development of autonomous

ships is more closely related to the topic of automated

train operation, because these entail even higher

individual costs for test vehicles and a comparable

availability and utilization of test sites as in railway

technology. Scaled ships are used as test vehicles, the

length of which is chosen to allow for a lighter

regulatory framework is possible in terms of driving

licenses and operation. Due to the large initial size of

ships, it is still possible to use original LiDAR,

camera and radar sensors even after scaling. (Kolewe

& Tietz, 2023).

In the automotive industry, the use of scaled

models for the development and testing of

autonomous driving functions is unknown. It is

assumed that the costs for test vehicles and test sites

are too low to consider investing in scaled models.

3.2 Scaled Models in Railways

In the field of rail vehicles, the use of scaled models

is already in use in teaching and research.

(Aceituno et al., 2017) shows the use of a scaled

vehicle with a 5-in track gauge for the validation of

dynamic simulations.

In the railway operations laboratory (ger.

“Eisenbahnbetrieblabor”) of TU Dresden a model

layout in H0 gauge (16.5 mm) is provided for,

teaching and research focused on safety technology

and signal box technology, with the possibility of

establishing a link to real systems such as the Dresden

suburban railway (TU Dresden, n.d.).

RWTH Aachen University also operates an H0-

scale environment featuring different realistic signal

box technologies, with the ability to simulate real

switch towers, as well as different kinds of safety

technologies up to ETCS simulation, and an extensive

track network to elaborate solutions for dispatch

optimization (RWTH Aachen, n.d.).

In the literature, scaled models are utilized in

railway technology for research, development and

demonstration of safety and interlocking technology,

timetable optimization and validation method for

dynamic simulation. An approach to establish scaled

model methods for on-sight ATO is described below.

4 SCALED MODEL

DEVELOPMENT FOR

ONSIGHT AUTOMATED TRAIN

OPERATION

To map on-sight ATO functions in the scaled model,

it is necessary to develop and set up an appropriate

test field and a demonstrator vehicle. The first step is

to clarify which functions are to be tested with the

Introducing Scaled Model Development to on-Sight Automatic Train Operation

381

development and testing platform. With reference to

a parallel project in which a shunting locomotive is

being automated, the test field and the demonstrator are

to be adapted to the needs of automated shunting

operations in order to improve the comparability of the

results. Due to the increasing complexity and

development effort required for GoA4 functions, it is

expected that scaled models can contribute to

development, especially for highly automated vehicles,

due to the advantages already described. Accordingly,

the scaled model should be planned in a modular way,

so that functions up to GoA4 can be developed.

4.1 Scaled Test Environment and

Demonstrator Locomotive

A marshalling yard usually contains arrival tracks, a

hump, directional tracks and classification tracks. The

tasks of a (automated) shunting locomotive are

divided into use cases for testing and development

purposes. These are “check, move, approach, attach,

pressing-up, push-loose, follow, closing gap, retrieve

wagon, clear track, emergency brake, buffing,

humping and move-up” (Hofmeier et al., 2022). For

an initial exploration of the possibilities of scaled

models for the development of ATO functions, the

focus will be limited to a reduced range of functions

of the shunting locomotive. For the experiment, the

use cases will include "move," where the locomotive

switches the track by moving via the dead-end track;

"approach" and "attach," where the locomotive

approaches an obstacle or wagon at a safe distance or

attaches to it; and "check," where the locomotive

verifies the functionality of the sensors. These use

cases are primarily performed in the area of the arrival

tracks and the hump; consequently, these must be

considered at least in the test field setup.

Figure 1: Track layout with arrival tracks (1), hump (2) and

dead-end track (3).

Figure 2: Overview of test area.

The test environment (Figure 1) is constructed on

a u-shaped plate with external dimensions of 6.2 m x

3.6 m, constrained by the room size (Figure. 2) and

features approximately 34 m of track length,

incorporating 17 switches and two three-way

switches, with a track gauge of 45 mm.

The test area consists a track harp as arrival tracks

(1), a hump with two parallel tracks (2) and a dead-

end track (3) used for switching between the tracks.

A range of rail vehicles are also available for

testing, including flat wagons, container wagons,

timber wagons and vehicle transporters.

In order to implement automated driving,

appropriate sensors must be positioned, data analyzed

and actuators controlled. In terms of sensor data, this

primarily requires data for environment detection and

localization. A comparatively high level of

computing power is expected to be required to

analyze environmental data. A Raspberry Pi appears

to be a sensible solution, as it also provides a large

number of interfaces. State-of-the-art ATO systems

mainly use LiDAR sensors for obstacle detection.

Accordingly, a LiDAR sensor should also be used in

the scaled model. It is possible to estimate the

installation space for a demonstrator from the

minimum requirements for computing power, sensors

and interfaces. The G or 1 gauge with 45mm (scale

1:32 to standard gauge) offers the possibility of a

sufficiently large demonstrator to accommodate all

the required components and keeps the effort and

costs for infrastructure construction to a minimum

compared to larger model railway gauges.

Figure 3: Demonstrator setup.

VEHITS 2024 - 10th International Conference on Vehicle Technology and Intelligent Transport Systems

382

Figure 3 shows the structure of the demonstrator

vehicle. The central unit is a Raspberry Pi 4B 8GB

(a). Two Arduino Nano (b) are used to read out sensor

data. The localization is implemented with an

ultrasound-based Marvelmind Indoor Positioning

System with mobile hedges on the vehicle (c) and

fixed beacons at the labs walls. To record odometry

data, an incremental encoder (d) is attached to the

engine shaft. A strain gauge is installed in (e), which

can be used to measure the trailer load. An RGB

camera RPI WWCAM (f), an ultrasonic sensor HC-

SR04 (g) and a LiDAR sensor RPLIDAR M2A8 (h)

are used as sensors for environment detection.

The sensor data from the strain gauge and

ultrasonic sensor are read out on Arduino 1, and data

can also be visualized on the LCD display via Arduino

1. The incremental encoder is processed with Arduino

2. Both Arduinos are connected to the Raspberry Pi

via I

C. The hedges of the indoor positioning system

are connected to the Raspberry Pi via USB.

In the next step, an ATO system is set up with

these components.

4.2 Developing a Scaled ATO Shunting

System

To fulfil the use cases defined at the beginning, the

first step is to derive the position of the locomotive on

the route or in the track plan using reliable positioning

in combination with the route data transmitted

externally to the locomotive (X-coordinates, Y-

coordinates, additional information). Additional

information includes, for example, speed

specifications between two waypoints and expected

obstacles such as a wagon on the track. The Indoor

Position System offers a 100 Hz protocol in which,

among other data, the X and Y coordinates of the

hedges are transmitted via USB. The ultrasonic

beacons use an IMU to interpolate between the

sampling rate of the actual ultrasonic system to

increase the data rate up to 100 Hz.

The localization data is read out via USB on the

Raspberry Pi and the route data is transferred via

TCP/IP.

The hardware support of Simulink for Raspberry

Pi is used for programming. The development process

here is analogue to the development on a dSPACE

MicroAutoBox II used in the real locomotive

application.

Figure 4 describes the procedure for determining

the position. Data is transferred to Simulink via WiFi,

USB and I²C interfaces. There, the position data is

checked for plausibility (verification of coordinates in

the

laboratory), then the measurement data is

Figure 4: Interfaces, process and components of

positioning.

projected onto the distance data in order to eliminate

the lateral deviation. The actual position is estimated

using the odometry data and a Kalman filter. It is also

possible to determine the locos orientation from the

positions of the hedges A & B.

This process enables the determination of the

locomotive's position on the route with sufficient

accuracy, sampling rate, and reproducibility.

The next step is to set up the environment

monitoring system. In the first approach, the

ultrasonic sensor is used to detect obstacles in order

to reduce complexity. This makes it possible to enable

platooning between two demonstrator vehicles in a

straight line. As it is not possible to clearly determine

where the object detected by the ultrasonic sensor is

located, this system cannot be used to ensure safe

movement in the track curve.

The LiDAR sensor is integrated into the system

for fully functional environment detection. For this

purpose, Robot Operating System (ROS) is installed

as a framework to provide LiDAR sensor data.

To make a driving decision based on the LiDAR

data, the raw data is transferred to Simulink via the

ROS topic and processed there in X/Y format

coordinates. In addition, the position of the sensor in

the world coordinate system is calculated using the

position of the beacons and the position of the LiDAR

on the locomotive (see Figure 5). By adding the

sensor position to the point cloud, the detected objects

are available in world coordinates.

Figure 5: Decision-making obstacle detection.

The distances of the objects to the track are

calculated taking the route data into account. If the

distance falls below a defined nominal value (e.g.

Introducing Scaled Model Development to on-Sight Automatic Train Operation

383

clearance profile), the object is classified as an

obstacle. If the distance to obstacles approaches the

stopping distance, braking is initiated.

Figure 6 shows the LiDAR point cloud after the

locomotive has approached. The sensor is installed in

such a way that the buffer level of the wagon is

intersected. Buffer and wagon front can be clearly

distinguished at a short distance.

Figure 6: Point cloud after approaching.

Figure 7 visualizes the ATO system in action. The

decision making is able to guide the loco along the

route. During the journey, the system checks the sensor

data for plausibility and executes the use cases check,

move, approach and attach with sufficient reliability.

Consequently, the implementation of ATO functions in

the model has been successfully completed.

Figure 7: Visualization LiDAR-position, route data and

point cloud data.

For the test runs of the “move” use case, target

positions are defined on the track to be approached by

the demonstrator. The system always determines the

correct orientation and therefore moves in the right

direction. The distance to the target point can be

determined at the time of measurement with a

maximum deviation of ±15 mm. In addition, the error

increases with speed (approx. 50 mm/s and effective

sampling rate (approx. 75 - 85 Hz) to the next

measurement point.

For the “approach” case, the system is able to

reliably determine the distance to obstacles within the

clearance gauge. The deviation of the distance

measurement is lower than that of the localization

system and therefore depends heavily on the local

localization accuracy. During the investigations,

borderline cases such as objects at the clearance

gauge boundary were excluded. It is expected that

further software (and hardware) improvements to the

system will be necessary in order to be able to

maintain tight tolerances (<10 mm) around the

boundary of the clearance gauge.

“move” and “approach” have in common that a

(defined) braking curve must be maintained, for

which precise speed information is required. The tests

show that the installed incremental encoder works

unreliably at times and that controlled braking is

intolerably often (>40 %) not possible. Furthermore,

it is not possible to model the realistic control

behavior of a real locomotive with this setup. A

revision of the speed sensor is necessary to increase

the functionality of the demonstrator.

The distances to be covered in the "attached" case

are short and can, in principle, be achieved with the

system stand. However, it is not possible to

implement a realistic approach speed behavior

depending on the load or other factors.

CONCLUSION

The test setup demonstrates that it is possible to

develop an ATO system in a scaled model in a cost-

and time-efficient manner. The sensors used in reality

can either also be scaled or have to be mapped using

corresponding substitution technologies (e.g. GPS 

ultrasonic positioning). Also, the processes and

programming languages used in reality can be

transferred directly to the model. The problems and

hurdles with regard to different interfaces and

latencies are analogous to reality. Accordingly, it is

expected that it will be possible to transfer the

knowledge gained to the real application.

The most important step is to demonstrate

scalability into the real system. Various hurdles are

expected here that could limit the usability of scaled

models. For instance, the effects caused by masses are

primarily distorted through scaling. Mechanical

processes may, in certain cases, fail to depict reality

with sufficient accuracy for estimating loads or similar

factors. Nevertheless, alternative solutions are already

delineated in the literature regarding this matter.

Moreover, it is anticipated that processes and structures

for environment detection and localization, as well as

interfaces and software architectures, can be

effectively developed within the demonstrated scaled

model approach. However, it is essential to identify

and determine dimensionless quantities for scaling the

individual system components to achieve this.

To this end, it is planned to expand the test field

structure in order to be able to consider further

application areas of driving on-sight and to generate

VEHITS 2024 - 10th International Conference on Vehicle Technology and Intelligent Transport Systems

384

a more flexible and versatile test environment that

requires increased performance in the system.

Additionally, the vehicle will be equipped with

enhanced sensors to address the described issues.

REFERENCES

Aerospace Engineering. (2022, September 15). Inside the

Wind Tunnel: How Aircraft Planes are Tested. https://

www.aerospacengineering.net/inside-the-wind-tunnel-

how-aircraft-planes-are-tested/

Atar, M. (2022). Scaled Earthquake Resistant Structures.

https://pure.manchester.ac.uk/ws/portalfiles/portal/213

188780/FULL_TEXT.PDF

DB Systemtechnik. (2021, September 30). Digitale S-Bahn

Hamburg—A safety concept to operate ATO over

ETCS. https://www.era.europa.eu/system/files/2022-

11/ control _major_risks_-_marc.pdf

European Commission. (2023). State of play—European

Commission. https://transport.ec.europa.eu/transport-

modes/rail/ertms/state-play_en

European Union. (2019). Digitalisation in railway transport.

https://www.europarl.europa.eu/RegData/etudes/BRIE/

2019/635528/EPRS_BRI(2019)635528_EN.pdf

Geurts, C., & Van Bentum, C. (2007). Wind Loading on

Buildings: Eurocode and Experimental Approach. In T.

Stathopoulos & C. C. Baniotopoulos (Hrsg.), Wind

Effects on Buildings and Design of Wind-Sensitive

Structures (Bd. 493, S. 31–65). Springer Vienna.

https://doi.org/10.1007/978-3-211-73076-8_2

Greiner-Fuchs, L., Schäfer, S., Hofmeier, T., & Cichon, M.

(2022). Database-supported methodical approach for

the development of a toolchain for the evaluation of

ATO functions using a scenario-based test

methodology. Proceedings of the Fifth International

Conference on Railway Technology: Research,

Development and Maintenance.

https://doi.org/10.4203/ccc.1.13.4

Guglielminetti, P., Lagraulet, M. F., Artuso, D., Lunadei,

E., Musso, A., Fusco, G., Licciardello, R., Piccioni, C.,

& Seno, F. (2015, Juli). Study on Single Wagonload

Traffic in Europe – challenges, prospects and policy

options. https://transport.ec.europa.eu/system/files/20

17-02/2015-07-swl-final-report.pdf

Hofmeier, T., Greiner-Fuchs, L., Schäfer, S., & Cichon, M.

(2022, Oktober). Task analysis of a shunting

locomotive to derive use-cases for scenario based tests

of ATO Functions. 9th Auto Test Conference, Stuttgart.

https://www.researchgate.net

IEC. (2006). IEC 62290-1:2006 Railway applications –

urban guided transport management and command/

control systems – part 1: System principles and

fundamental concepts (IEC 62290–1).

Kaack, L. H., Vasihnav, P., Morgan, M. G., Azevedo, I. L.,

& Rai, S. (2018). Decarbonizing intraregional freight

systems with a focus on modal shift. Environmental

Research Letters, 13(8). https://doi.org/10.1088/1748-

9326/aad56c

Kolewe, B., & Tietz, L. (2023, Oktober 5). Vom Seehafen

bis zur Haustür – autonomer Gütertransport auf dem

Wasserweg. Fachkonferenz "Neue Technologien und

Testfelder für Wasserstraßen und Häfen.

Pachl, J. (2020). Railway Signalling Principles.

Universitätsbibliothek Braunschweig. https://doi.org/

10.24355/ DBBS.084-202006161443-0

Pegasus. (2019). PEGASUS Method: An Overview.

https://www.pegasusprojekt.de

Pourian, T. N. (2023). Effects of Automation in Railway on

the Capacity – At the Example of the S-Bahn Stuttgart

https://www.diva-portal.org/smash/get/diva2:1777801/

FULLTEXT01.pdf

Powell, J. P., Fraszczyk, A., Cheong, C. N., & Yeung, H. K.

(2016). Potential Benefits and Obstacles of

Implementing Driverless Train Operation on the Tyne

and Wear Metro: A Simulation Exercise. Urban Rail

Transit, 2(3–4), 114–127. https://doi.org/10.1007/

s40864-016-0046-9

Rio Tinto. (2017, Oktober 2). Rio Tinto completes first fully

autonomous rail journey in Western Australia.

https://www.riotinto.com/en/news/

RLC Wustermark. (2023, Juni 21). Zugangs- und

Nutzungsbedingungen. Rail & Logistik Center

Wustermark GmbH. https://www.rlcw.de/zugangs-und-

nutzungsbedingungen.html

RWTH Aachen. (n.d.). Die Eisenbahntechnische Lehr-

Versuchsanlage der RWTH Aachen. https://www.via.

rwth-aachen.de/elva.php

Schäfer, S., Greiner-Fuchs, L., Hofmeier, T., & Cichon, M.

(2023). Entwicklung eines echtzeitfähigen virtuellen

Laborprüfstands (Simulationsumgebung) für das

szenariobasierte Testen und Validieren hoch-

automatisierter Fahrentscheidungs- und Steuerungs-

systeme von Rangierlokomotiven. Tagungsband 19.

Internationale Schienenfahrzeugtagung Dresden,

Tagungsband, 108.

Schnieder, L. (2021). European Train Control System

(ETCS). Springer Berlin Heidelberg. https://doi.org/

10.1007/978-3-662-62878-2

Tagiew, R., Leinhos, D., von der Haar, H., Klotz, C., Sprute,

D., Ziehn, J., Schmelter, A., Witte, S., & Klasek, P.

(2022). Onboard Sensor Systems for Automatic Train

Operation. In S. Marrone, M. De Sanctis, I. Kocsis, R.

Adler, R. Hawkins, P. Schleiß, S. Marrone, R. Nardone,

F. Flammini, & V. Vittorini (Hrsg.), Dependable

Computing – EDCC 2022 Workshops (S. 139–150).

Springer International Publishing.

Trepáčová, M., Kurečková, V., Zámečník, P., & Řezáč, P.

(2020). Advantages and disadvantages of rail

transportation as perceived by passengers: A qualitative

and quantitative study in the Czech Republic.

Transactions on Transport Sciences, 11(3), 52–62.

https://doi.org/10.5507/tots.2020.014

TU Dresden. (n.d.). The Laboratory. TU Dresden. https://tu-

dresden.de/bu/verkehr/ibv/ebl/das-labor/start

seite?set_language=en

Zasiadko. (2019, November 19). Fully automated metros

run in six EU countries. RailTech.Com.

https://www.railtech.com/infrastructure

Introducing Scaled Model Development to on-Sight Automatic Train Operation

385