You are on page 1of 11

5G for remote driving of trains

Yamen Alsaba1[0000−0001−6165−3265] , Marion Berbineau2,1[0000−0003−3807−9669] ,


Iyad Dayoub3,1 , Emilie Masson1 , Gemma Morral Adell4 , and Eric Robert5
1
Railenium, yamen.alsaba.railenium.eu
2
COSYS, University Gustave Eiffel, F-59650 Villeneuve d’Ascq-Frannce
marion.berbineau@uni-eiffel.fr
3
UPHF, iyad.dayoub@uphf.fr
4
SNCF, gemma.morral-adell@sncf.fr
5
Thales GT, eric.robert@thalesgroup.com

Abstract. Automatic Train Operation (ATO) is a new growing mar-


ket in the railways sector since 2019. Several railways operators such
as SNCF, DB, SBB and so on have launched deep transformation in
their infrastructures and rolling stock in order to enter in digital era.
One of this game changers is upgrading to autonomous train on exist-
ing or new infrastructure. Since autonomous train means the absence
of train driver, there is a big need of uplinked information to supervise
autonomous trains from the ground. This includes the need of remotely
driving the train if it encounters a problem, e.g. a non-recognised obsta-
cle, an infrastructure breakdown. Remote driving will be a new operation
mode in the railways sector that will rely on a well-designed radio link
provided by 5G. This paper presents preliminary results on test tracks
with Long Term Evolution (LTE) and a simulation based comparison
between LTE and 5G at physical layer using Non orthogonal Multiple
Access.

Keywords: autonomous trains, remote driving, 5G, OFDM, NOMA

1 Introduction
Full automation of trains will allow increasing drastically infrastructure capac-
ity, optimizing train operations in general and also speed of the trains. Safety,
security and passenger service issues are also targeted namely punctuality, train
according to demand, etc. Driverless system already exists in the urban segment
with full automation of train operation on dedicated lines. The next challenge,
is now to generalize automation to other railway segment such as freight, re-
gional and main lines with possibly mix of traffic between trains with driver and
driverless trains. In this context, a mandatory brick relies on the remote driving
of trains.
The demonstration 6 and development of such a brick is the aim of the TC-
Rail project, a partnership formed by SNCF, Thales, Actia Telecom, CNES and
6
TC-RAIL demo: https://www.sncf.com/fr/groupe/newsroom/teleconduite-train-
autonome [Last accessed 12th July 2020]
2 Y. Alsaba et al.

Railenium, that will demonstrate the possibility of driving a locomotive safely


from a remote location, without a driver in the train cabin, with a level of
safety similar to that obtained in presence of a driver in the train. This project
constitutes the first proof of concept for telecontrol of a train without European
Railway Management System (ERTMS) infrastructure and at maximum target
speed of 100 km/h. It is foreseen to remove the main technical obstacles that
could prevent such exploitation.
As the driver is no longer in the cabin, video of what the train perceives
in front of it associated with other perception information is transmitted to a
distant site where is the driver, in what so called ”the eyes of the train”. The
train-to-ground transmission link will therefore have to be very high data rate
with a high quality of service so that the remote driver can have a vision similar
to that which he would have if he were in the cabin of the train. Consequently,
remote driving is based on three major technological blocks: a good perception
system able to combine video with other information (audio, localization, etc.), a
high data rate, robust, reliable and ultra-low latency radio communication uplink
between the train and the remote site and a remote driving Human Machine
Interface (HMI) with ergonomics suitable for a new job position in railways
sector, i.e. remote train driver. There are three main applications for remote
driving of trains:
• the management of sectors between yard and the client’s site called “last
kilometers” to reduce prolonged periods of transportation and waiting time
for the drivers;
• the management of technical routes between maintenance centers and sta-
tions;
• the recovering from an autonomous train (failing or not).
In this paper we focus on the wireless link of the remote driving of the train.
The rest of the paper is organized as follows: Section 1 gives a brief state of
the art related to wireless communication for railways. Section 2 summarises the
results obtained during the preliminary experiments performed with 4G/ LTE.
In Section 3, we propose a comparison between LTE and 5G at physical layer
using Non Orthogonal Multiple Access (NOMA) in the case of two trains in the
cell. Finally we conclude and give perspectives.

2 Wireless communications for railways


GSM-R is used in the European railway sector for the control and command of
high speed trains. Based on GSM phase2+ system, its date of obsolescence is
predicted by 2030. To anticipate, a new system called Future Railway Mobile
Communication System (FRMCS) is under development and preliminary specifi-
cations have been published [8]. The most important and mandatory characteris-
tics of this new system are: Internet Protocol (IP) based communications, bearer
agnostic, flexibility and resilience to technological evolution. FRMCS should sat-
isfy all the needs for existing critical communications but also new ones related
5G for remote driving of trains 3

to driverless trains, virtual coupling and decoupling of trains, and communica-


tions with sensors along the tracks. In addition, non-critical applications such
as real time video calls, augmented reality data communication, and wireless
internet on-train for passengers, should be supported [11]. The Long Term Evo-
lution (LTE) system has been particularly studied [6, 9] as a serious candidate
to replace GSM-R associated with other radio access technologies (RAT). How-
ever, the fact that FRMCS requires higher data rate and higher bandwidths
due to real time HD video transmission for remote driving of train for example,
has pushed researchers and industry to envision the 5G wireless communication
system as another alternative [1, 12]. Furthermore, in the European Shift2rail
program, prototypes of a new Adaptable Communication System (ACS) are un-
der development by industry [3]. The aim is to combine different RAT e.g., 4G,
5G, Wi-Fi and Satellite communications that will cooperate to provide the re-
quired communication needs of the different safety and non safety related railway
applications.

3 Preliminary experiments on the tracks


In the framework of the TC-Rail project, an LTE infrastructure in Time Divi-
sion Duplex (TDD) mode was deployed specifically on a small area (around 4
km) covering a portion of a french line in Paris region in order to evaluate the
performance of this dedicated technology for the remote driving of the train.
It was deployed using eNodeB products from Nokia at 2.6 GHz with 20 MHz
bandwidth. The masts were located at 15 m above the ground level and near the
tracks. The preliminary tests were done in Single Input Single Output (SISO)
configuration. The maximal transmitted power was 43 dBm and the antennas
offered a gain of 16 dB. The radio coverage was performed by SNCF Réseau
teams along the tracks in order to optimize connectivity along the trip. The
measurements were done using a non GBR (Guaranteed Bit Rate) bearer with
a QCI (QoS Class Identifier) equal to 7. The results showed an average data
rate of 7.7 Mbits/s in uplink in the covered area. This is very satisfying as the
minimal requirement was 2 Mbps for video. The round trip latency was 60 ms
in average. Fig. 1 shows the train used for the demonstrator equipped with
cameras, antennas and modem, and the remote driver cabin developed in the
project.

4 5G and LTE Comparison


4.1 Context
Under the umbrella of the 5G and beyond wireless communication systems,
many enabling technologies have emerged recently that offers different improve-
ments to 4G systems [2]. Among them, a first enhancement concerns the high
system throughput offered by the Non-Orthogonal Multiple Access technology
(NOMA). In multiple users scenarios, the users access the radio network by
4 Y. Alsaba et al.

(a) Remote driving cabin (b) Testing train on the tracks

Fig. 1: The TC-Rail experimentation

sharing the available time and frequency resources. Conventionally, users will
be allocated different time and frequency resources in an orthogonal manner
such as in the Time Division Multiple Access (TDMA) and the Frequency Divi-
sion Multiple Access (FDMA) techniques. Recently, NOMA technology [7] has
gained a widespread interest due to its accompanied gains in the overall system
throughput. NOMA technique allows users to share the same time and frequency
resources by adopting a superposition coding schemes at the transmitter and a
successive interference cancellation schemes at the receiver. This proved to bring
critical improvements in terms of the achieved sumrate of the corresponding users
at the cost of additional interference components that needs to be considered
within the transceiver design process.
NOMA has been recognized as the potential multiple access scheme for future
communication systems. By virtue of exploiting power domain, NOMA can serve
multiple users at the same time, frequency, and code resources yielding higher
spectral efficiency. NOMA communication system implementation involves two
major processes namely Superposition Coding (SC) and Successive Interference
Cancellation (SIC) at the base station and users terminals, respectively. NOMA
users are distinguished according to their channel status, wherein users are al-
located with portion of power inversely proportional to their channel condition.
To decode their own messages, NOMA users suppress the information messages
of all weaker users, while considering the information of the stronger users as
interference.
Most of the literature on NOMA based communication systems has consid-
ered Perfect SIC (PSIC) process, i.e. an accurate knowledge of all weaker users
information messages is available at the each user’s terminal. However, this as-
sumption is not practical in the TC-Rail project as it implies that the user should
perfectly estimate both the amplitude of all weaker users waveform [5]. More-
over, this task becomes extremely challenging in doubly selective channels such
those encountered for vehicular and railway wireless communication systems.
A few literature can be found on imperfect SIC based NOMA system. NOMA
versus OMA based systems’ performance comparison has been carried out in
5G for remote driving of trains 5

the literature for different scenarios. NOMA superiority over OMA is proved in
terms of users fairness [10], multi-user capacity [15], beamforming aspects [4],
and cell-edge user data rate [14]. Numerical simulations illustrate that NOMA
scheme provides higher data-rate, higher spectral efficiency, lower latency. How-
ever, NOMA users suffer from inter-user interference.
To be compliant with the 3GPP 5G Phase 2 (release 16) [16] that adopts
NOMA as the potential multiple access candidate for the 5G systems, NOMA has
been adopted as the multiple access scheme for the TC-Rail project. It is worth
mentioning that Orthogonal Frequency Division Multiple Access (OFDMA) has
been adopted in the LTE and 3GPP standard release 8 as the multiple access
method.
In order to illustrate the gain provided by the proposed 5G-based communica-
tion systems, a comparison with LTE-based system is carried out at the physical
layer only, where no higher layers techniques are involved. As the 5G physical
implementation is not realized yet, and important parameters such as carrier
frequency and bandwidth are not identified especially in the railway system, the
TC-Rail implementation configuration and physical parameters are adopted to
perform the comparison. Table 1 illustrates the adopted setup configuration in
the comparison.

Table 1: TC-Rail Setup Configuration Parameters


Parameter Value
Carrier Frequency 2585 MHz
Bandwidth 20 MHz
Transmit Power 20 watts
Antenna Gain 16.5 dBi
Number of OFDM Subcarriers 1200
OFDM Subcarrier Spacing 15 kHz
Trains Velocity 100 km/h
Trains number 2
Channel Model and Doppler Vehicular A, Jakes
Monte-Carlo Simulation Realization Number 1000

The physical layer technologies in the LTE setup are 3GPP release 8 compli-
ant, wherein the OFDMA, OFDM, turbo coding are the technologies used for
the multiple access, waveforms and channel coding blocks respectively. OFDM
is adopted for both LTE and 5G communication systems. However, the OFDM
parameters in terms of number of subcarriers, subcarriers spacing, symbol and
CP length are different in the 5G than its values in the LTE, as the bandwidth
in the 5G is 100 MHz for operating frequencies below 6 GHz and 400 MHz at
28 GHz and above where the bandwidth is 20 MHz in the LTE. As we consider
the TC-Rail LTE-based setup and for the sake of fairness, the OFDM parameters
are kept the same in the LTE and 5G systems. Furthermore, the same MIMO
6 Y. Alsaba et al.

(Multiple Input Multiple Output) scheme is used for LTE and 5G systems to
guarantee the comparison to be fair as choosing different MIMO schemes will
change radically the system performance in terms of throughput and Bit error
rate (BER). Table 2 summarizes the technology used for both LTE and 5G in
the comparison.

Table 2: 5G and LTE Technologies


Technology LTE 5G
Multiple Access OFDMA TDD NOMA
Waveform OFDM OFDM
MIMO 2*2 Alamouti 2*2 Alamouti
Channel Coding Turbo Coding LDPC

The simulation involves two served trains at the same time, a train with good
channel condition (center train) and the second one with poor channel condition
(edge train). The comparison between the LTE and 5G is carried out in terms
of sum rate (the sum of the both trains rate) and BER with considering the
simulation parameters and technologies illustrated in tables 1 and 2.

4.2 System Model


We consider a communication system, wherein a base station is communicating
with two moving users at the center and the edge of the cell. Due to the mobility
of the users, doubly selective fading channel model is adopted. The transmitted
users’ messages are first mapped into a 2-dimensional space “a time frequency
space”, then transformed into the signal space via the synthesis function gm,k (t).
Hence, the transmitted signal can be expressed as follows:

XM
K−1 −1
X
s(t) = gm,k (t)xm,k (1)
k=0 m=0

where xm,k is the transmitted message at the mth subcarrier and the kth time
domain symbol .K denotes the number of time domain symbols and M is the
number of subcarriers of the whole transmission block. The synthesis function
gm,k (t) that maps xm,k into the signal space can be written as follows:

gm,k (t) = ptx (t − kT )ej2πmF (t−kT ) (2)

where ptx (t) is the pulse shape, also known as the prototype filter. This pulse
shape will determine the energy distribution (in time and frequency domains)
of the transmitted symbol. T is the symbol duration while F is the subcarrier
spacing. Hence, we can read Eq. 2 as follows: gm,k (t) is considered as the proto-
type filter ptx (t) with translation of kT and modulation of mF .
5G for remote driving of trains 7

After conducting the sampling process, Eq. (1) be represented in matrix form
s = Gx as [13], where

G = [g1,1 g2,1 ...gM,1 g1,2 ...gM,K ] (3)


T
x = [x1,1 x2,1 ...xM,1 x1,2 ...xM,K ] (4)
By sampling gm,k (t) in Eq. (2), the samples are grouped in one vector gm,k ∈
CN ×1 where gm,k = G(:, mk). N denotes the number of samples of the whole
transmission block.

In OFDM-based NOMA techniques, super-positioned coding is applied to


send edge and cell users symbols while sharing the same time and frequency
resources. In other words, at the mth subcarrier and the kth time symbol, the
sent message is written as follows:
√ √
xm,k = αdedgem,k + 1 − αdcenterm,k (5)

dedgem,k , dcenterm,k ∈ C are the transmitted symbols of the edge and the cen-
ter user, respectively. dm,k ∈ R is the possible special case e.g., Pulse Amplitude
Modulation (PAM). α and 1 − α are the power allocation factor for NOMA edge
and
 center user respectively and hence the transmitted power is normalized; i.e,
E |xm,k |2 = 1.

In the OFDM-based OMA case, different time and frequency resources are
allocated to the center and edge users, where
(
dedgem,k , (m, k) ∈ Ωedge
xm,k = (6)
dcenterm,k , (m, k) ∈ Ωcenter
We consider a fair distribution of resources among users, i.e. |Ωedge | =
|Ωcenter | = M K/2. where |Ω| indicates the number of elements in the set Ω.
At users terminals, the demodulated signal can be written as follows:

y = QHG x + Qη (7)
| {z }
STM

where STM = QHG is the corresponding System Transmission Matrix. The


non-diagonal elements of this STM represents the interference components, while
the desired signal dwells its diagonal elements. In the NOMA case, inter-user
interference exists even in the diagonal elements. However, at the center user,
SIC is implemented to remove the effects of the edge user interference.

4.3 Sum rate


The sum rate, which represents the overall data rate at both edge and center
users, is written as follows: Rsum = Redge + Rcenter where Redge , Rcenter express
the data rate of the edge and the center user, respectively.
8 Y. Alsaba et al.

NOMA case: we assume that each time frequency resource has the same data
rate. Hence, the data rate at user u ∈ {edge, center} is given as:

M
Ru = γ log2 (1 + SINRu ) (8)
T
In Quadrature Amplitude Modulation (QAM) based waveforms and for OFDM,
we give γ = 1. In order to calculate the SINR of the ith symbol, we start by
writing the ith row of the corresponding STM as follows

L−1
!H
X
H H
hH

D(i, :) = Q(i, :)HG = G diag qi,l l (9)
l=0

where diag(x) is the diagonal matrix


 of the vector x. We calculate the covariance
matrix C = E D(i, :)H D(i, :) as:

L−1
!
X
C = GH Rql Rhl G where H
Rql = qi,l qi,l (10)
l=0

At the edge user: we first need to calculate the power of the useful signal
Pedge = αC(i, i) , the power of the inter-user interference PInter = (1 − α)tr{C},
the power of the intra-user interference PIntra = α (tr{C} − C(i, i)) and hence:

Pedge
SINRedge = (11)
PIntra + PInter + Pn
where Pn is the noise power.

At the center user: the SIC process will eliminate the inter user interference
component, we write Pcenter = (1−α)C(i, i) and PIntra = (1−α) (tr{C} − C(i, i)),
and hence:

Pcenter
SINRcenter = (12)
PIntra + Pn

OMA case: in the OMA case, users won’t share same time and frequency
resources. By assuming fair distribution of resources among users, the bit rate
at each user is given as:

M
ROM A = γ log2 1 + SINROMA

(13)
2T

where the prelog factor of 1/2 in the rate equation is due to the fair resource
allocation between the two users. This implies also that PnOMA = (1/2)PnNOMA
5G for remote driving of trains 9

4.4 Simulation results

Fig. 2a illustrates the sumrate of the proposed 5G communication system and


LTE for different waveforms. The rate is almost doubled in the 5G case with
two trains due to the non-orthogonal resource allocation in 5G NOMA while
the trains need to share the time and frequency resources in LTE. However,
NOMA users suffer from inter-user interference that makes the gain in sumrate
between 5G and LTE nonlinear. The curves shown in Fig. 2b represents the
BER performance for different multi-carrier waveforms in LTE and 5G commu-
nication systems, namely OFDM, Filter Bank Multi Carrier (FBMC), Filtered
OFDM (FOFDM) and Weighted Overlap and Add (WOLA). The BER of all
waveforms in the 5G is better than that of the LTE. This is due to the fact
that non-orthogonal resources allocation in 5G is more robust toward chan-
nel impairments especially in the railway environment, where the orthogonality
doesn’t hold in the selective fading channel resulting from the mobility. Fig. 3
draws a 3D representation of the system sumrate in LTE and 5G system as a
function of both trains’ velocity. We can notice from the figure that the 5G sys-
tem maintains its superiority over LTE even in high speed regime. The sumrate
in 5G varies with speed between 180 and 160 Mbps at the highest speed for both
trains, however the sumrate in the LTE case is between 100 and 80 Mbps. Hence,
the 5G system is more robust against speed and the resulting non orthogonality
in the channel.

(a) Sumrate of LTE and 5G systems (b) BER of LTE and 5G systems

Fig. 2: Sumrate and BER results for LTE and 5G

5 Conclusion and perspectives

Trains are entering the era of full automation thanks to sensors and wireless
communications shifting control functions from the human driver to computers.
10 Y. Alsaba et al.

Fig. 3: Sumrate of LTE and 5G systems

Driverless systems already exist for metro and dedicated lines. The full automa-
tion of trains in the context of existing lines with the possibility to cross other
non automatic trains is very complex. To reach this challenge, a mandatory
brick is the remote control of a driverless train from a distant site thanks to
radio transmission. This will allow telecontrol of the train anywhere at any time
for example for specific maneuver in stations or marshalling yards or in case of
failure of the driverless system. The TC-Rail project aims to bring a proof of
concept of the remote control of the train. In this paper we have briefly pre-
sented the evolution of the wireless communication systems for trains and the
first performance results for the train-to-ground video transmission considering
LTE deployment along the line. Thanks to numerical simulations, we have com-
pared LTE and 5G performances at physical layer with the same characteristics
in the case of two trains in the cell and we have highlighted the importance to
consider NOMA techniques associated with OFDM and MIMO to guarantee a
good performances for both trains even in doubly selective channel and with
high speed condition.

Acknowledgements
This work has been carried out in the framework of the TC-Rail project co-
financed by a public and private consortium (Railenium, SNCF, Thales, Actia
Telecom, CNES).

References

1. Ai, B., Guan, K., Rupp, M., Kurner, T., Cheng, X., Yin, X.F., Wang, Q., Ma, G.Y.,
Li, Y., Xiong, L., et al.: Future railway services-oriented mobile communications
network. IEEE Communications Magazine 53(10), 78–85 (2015)
2. Akyildiz, I.F., Nie, S., Lin, S.C., Chandrasekaran, M.: 5G roadmap: 10 key enabling
technologies. Computer Networks 106, 17–48 (2016)
5G for remote driving of trains 11

3. Allen, B., Eschbach, B., Mikulandra, M.: Defining an adaptable communications


system for all railways. In: proceedings of the 7th Transport Research Arena TRA
2018 (TRA (2018)
4. Alsaba, Y., et al.: Full-duplex cooperative non-orthogonal multiple access with
beamforming and energy harvesting. IEEE Access 6, 19726–19738 (2018)
5. Andrews, J.G., et al.: Optimum power control for successive interference cancella-
tion with imperfect channel estimation. IEEE Transactions on Wireless Commu-
nications 2(2), 375–383 (2003)
6. Brunel, L., Bonneville, H., Charaf, A.: Throughput performance of 3GPP LTE
system in railway environment. In: International Workshop on Communication
Technologies for Vehicles. pp. 60–71. Springer (2017)
7. Cai, Y., Qin, Z., Cui, F., Li, G.Y., McCann, J.A.: Modulation and multiple ac-
cess for 5G networks. IEEE Communications Surveys Tutorials 20(1), 629–646
(Firstquarter 2018). https://doi.org/10.1109/COMST.2017.2766698
8. FRMCS Functional Working Group: Future Railway Mobile Communication Sys-
tem, User Requirements Specification, FU-7100 (January 2019)
9. He, R., Ai, B., Wang, G., Guan, K., Zhong, Z., Molisch, A.F., Briso-Rodriguez,
C., Oestges, C.P.: High-speed railway communications: From GSM-R to LTE-R.
IEEE Vehicular Technology Magazine 11(3), 49–58 (2016)
10. Li, A., et al.: Investigation on low complexity power assignment method and perfor-
mance gain of non-orthogonal multiple access systems. IEICE trans. Fundamentals
97(1) (2014)
11. Moreno, J., Riera, J.M., d. Haro, L., Rodriguez, C.: A survey on
future railway radio communications services: challenges and opportu-
nities. IEEE Communications Magazine 53(10), 62–68 (October 2015).
https://doi.org/10.1109/MCOM.2015.7295465
12. Mottier, D.: How 5g technologies could benefit to the railway sector: challenges
and opportunities. In: Mitsubishi Electric R&D Centre Europe–France (2018)
13. Nissel, R., et al.: Filter bank multicarrier modulation schemes for future mobile
communications. IEEE Journal on Selected Areas in Communications 35(8), 1768–
1782 (2017)
14. Saito, Y., et al.: Non-orthogonal multiple access (NOMA) for cellular future radio
access. In: 2013 IEEE 77th vehicular technology conference (VTC Spring). pp. 1–5.
IEEE (2013)
15. Weingarten, H., et al.: The capacity region of the Gaussian multiple-input multiple-
output broadcast channel. IEEE transactions on information theory 52(9), 3936–
3964 (2006)
16. Yuan, Y., Yuan, Z., Tian, L.: 5g non-orthogonal multiple access study in 3gpp.
IEEE Communications Magazine 58(7), 90–96 (2020)

You might also like