© 2024 IEEE. Personal use of this material is permitted.

Permission from IEEE

must be obtained for all other uses, in any current or future media,
including reprinting / republishing this material for advertising or promotional
purposes, creating new collective works, for resale or redistribution to
server or lists, or reuse of any copyrighted component of this work in other
works.
Feasibility of Direct Air to Ground Communication
via a Terrestrial 5G Network
Syed Aizaz Ali Shah∗ , Thomas Meyerhoff† , Rainer Grünheid∗ , Gerhard Bauch∗ and Dominic Schupke†
∗ Institute
of Communications, Hamburg University of Technology, Hamburg, Germany.
† Central
Research and Technology, Airbus Operations GmbH, Hamburg/Munich, Germany.
Email: [email protected], [email protected], {gruenheid, bauch}@tuhh.de, [email protected]

Abstract—This feasibility study deals with Air to Ground networks is also explored, for example, in [7]–[9]. The support
(ATG) communications between a 5G type Terrestrial Mobile of aUEs onboard UAVs by 5G networks is already a part of
Communication Network (TMCN) and an User Equipment the standardization [10].
(UE) installed onboard a passenger aircraft in approach to an
airport. The network model is based on real gNodeB (gNB) Cellular connectivity for high altitude aUEs, such as aircraft,
deployment data and optimized for serving terrestrial UEs. The is also of interest. The European Aviation Network (EAN) is
evaluation concerns a scenario in which the passenger aircraft is an ATG communication system that is offering seamless in-
supported by a remote operator via an ATG link to assist its safe flight broadband connectivity for commercial passenger air-
landing. To mitigate the interference expected to arise in ATG
craft via satellites and dedicated, sparsely deployed LTE Base
communications, the aircraft is assumed to be equipped with
an Active Electronically Scanned Array (AESA). The simulation Stations (BSs) with mechanically up-tilted antennas covering
results discussed in this paper address the Signal to Interference the airspace [11]. Efforts to provide broadband connectivity
power Ratio (SIR), link availability, and handover process during to passengers onboard aircraft via 5G networks are already
ATG communications along representative approach trajectories underway [12].
for different AESA configurations and levels of support by the
TMCN. It is shown that SIR above 30 dB and link availability
ATG communication with an aUE is subject to strong
upto 97.7% can be achieved by using an 8×8 AESA at the aircraft interference in both up- and downlink [2]–[9]. Network based
when the network provides support in the form of dedicated solutions to mitigate the interference include partitioning of the
radio resources. It is observed that the frequency of handovers radio resources between terrestrial and aUEs, inter-cell inter-
increases with improving link quality. ference coordination methods and power control mechanisms
I. I NTRODUCTION [4], [6], [13], [14]. The use of directional antennas on UAVs
has been identified as a key aUE based solution to combat
In aviation, the concept of remote operation is typically
interference [2], [14]–[18]. The use of an AESA onboard an
associated with Uncrewed Aerial Vehicles (UAVs). However,
UAV that is capable of tracking a serving BS with its beam
it may also be applied to passenger aircraft in the future.
can significantly reduce the rate of handovers and probability
One possible use case could be to improve flight safety by
of communication outage [18].
enabling a remote operator to assist the landing of an aircraft
Contrary to the EAN and 5G related ATG standardization
e.g. in an emergency. In this case, wireless transmission of
activities for high altitude aUEs [12], this work concerns ATG
Command and Control (C2) information between an aircraft
communications for C2 links. We assume a case of emergency
and the remote operator will be required. Consequently, the
requiring safe landing of the aircraft at the nearest airport,
C2 link will be subjected to strict availability, reliability and
which must be supported by the remote operator. The results
latency requirements [1].
presented hereafter give a first reference to the feasibility of
The ubiquity of TMCNs such as LTE and 5G and their
using an existing TMCN in this context and illustrate the
infrastructure make these networks promising candidates for
challenges arising from the use of infrastructure optimized for
connecting a remote operator to an aircraft. In that context,
terrestrial coverage.
an UE onboard aerial vehicles is referred to as Aerial User
Equipment (aUE) and the communication between an aUE
II. A IR TO G ROUND C OMMUNICATION M ODEL
and a TMCN is referred to as ATG communications.
ATG communications have been studied intensively in the The ATG communication model considered in this study is
context of beyond line of sight communications for UAVs depicted in Fig. 1 and consists of:
flying at low altitudes. Field test and simulation results for • a 5G TMCN with gNBs equipped with down-tilted sector
aUE moving over LTE networks at a height of up to 300 m are antennas, i.e., primarily intended for terrestrial users,
discussed in [2]–[6]. The results, e.g., in [6], indicate that the • an aUE installed at an aircraft on approach trajectory,
downlink quality should be sufficient for C2 links at altitudes • terrestrial UEs and
up to 120 m. ATG communications between UAVs and 5G • a Ground Operations Center (GOC) that is facilitating a
The authors would like to thank Stefan Neumann (Hamburg University of remote operator with a wired communication link to the
Technology) for assisting with the simulations. TMCN.
 Uplink 150◦ 30◦
Downlink 75

270◦

25
GOC

y in km
Fig. 1. ATG communications towards an aircraft in approach
-25
The aUE communicates with a serving gNB on up- and
downlink. The serving gNB relays the data to the GOC through
GSw : Swiss gNBs Rome
a wired backhaul. The remote operator is located in the GOC
-75 GFo : Foreign gNBs Venice
and able to control the aircraft via a C2 link. In this regard, it Philadelphia
is expected that an aUE onboard an aircraft in approach:
• is in radio range to many gNBs,
-125 -75 -25 25 75 125
• is predominantly served via the side lobes of the down- x in km
tilted sector antennas, Fig. 2. 5G network for rural coverage in Switzerland along with foreign 5G
network in neighboring countries within 100 km of Zurich airport.
• experiences severe interference on the downlink and
• causes interference on the uplink of terrestrial UEs.
in the neighboring countries. A foreign rural 5G network
Similar to the case of UAVs, the challenges for ATG
composed of a set GFo of gNBs is considered within 100 km
communication with an aircraft in approach are to maintain
radius of Zurich airport, depicted by red dots in Fig. 2.
a sufficient link budget towards the serving gNB and mitigate
The deployment of the foreign gNBs is estimated, based on
interference. However, for the latter, the interference effects are
public data of 5G coverage in the neighboring countries1 . To
expected to be more severe compared to the case of UAVs, as
facilitate ATG communications during the complete approach,
there will be significantly more gNBs in radio range due to
a coverage area of 50 km radius around the Zurich airport (blue
the high altitude of the aUE.
circle in Fig. 2) is considered.
In this work, the feasibility of ATG communications for
aircraft in approach is studied with simulations on link budget
B. Aerial User Equipment and gNodeB
level. The study focuses on the downlink only because the
anticipated strong interference poses a greater challenge for The aUE is installed onboard a passenger aircraft and
an aUE as compared to terrestrial UEs. It is assumed that connected to a square AESA. The AESA is integrated into
interference caused by an aUE in an emergency situation is the underside of the fuselage with its boresight pointed down-
tolerated. wards, as indicated in Fig. 1. The antenna array consists of
The simulation is based on the following model blocks: Ne elements, spaced at 0.5λ, in both horizontal and vertical
directions where λ is the carrier wavelength in meters. The
A. Terrestrial Mobile Communication Network
individual array elements are modeled according to [20].
The terrestrial network model is composed of gNBs with Further, the aircraft orientation is constrained such that its roll
high transmit power levels, low operating frequencies and and pitch axis remain parallel to the earth surface during its
high antenna masts. These properties can be considered as flight.
advantageous for ATG communications and are typical for Each gNB is modeled with a mast height of hgNB =35 m
gNBs intended for rural coverage. and three sector antennas pointed towards 30◦ , 150◦ and 270◦
To create the network model, real deployment information azimuth (top left of Fig. 2). Each sector antenna is a fixed beam
of gNBs in the vicinity of the Zurich international airport 8×1 vertical Uniform Linear Array (ULA) with 0.5λ element
was obtained from the data provided by [19]. The data set spacing and has a mechanical down-tilt of 3◦ [21]. The array
contains the gNB locations and their classification into four elements are modeled according to [20]. A transmit power of
categories based on their total radiated powers. Therefore, the 46 dBm is assumed for each cell. Relevant parameters of the
network model was created from the data set in two steps: gNB and aUE models are summarized in Table I.
First, only the gNBs that belong to the category of highest
total radiated power are considered. Second, inspired from the C. Link Budget
description of rural macro scenario in [20], a minimum Inter-
site Distance (ISD) of 1, 732 m is enforced since a mobile The average received power observed by the aUE from any
operator will not install these type of gNBs very close to each gNB sector is computed as:
other. The resulting network comprised by the set GSw of gNBs P̄R,dBm = PT,dBm + GT,dBi + GR,dBi − LdB , (1)
is depicted in Fig. 2 by the grey dots.
Due to close proximity of the Zurich airport to international 1 https://www.telekom.de/netz/mobilfunk-netzausbau, https://www.magenta
borders, it is vital to consider the presence of 5G networks .at/unternehmen/netz and https://www.nperf.com/en/map/5g
 TABLE I TABLE II
gNB AND aUE PARAMETERS C HANNEL MODEL FOR hUE ≤ 300 m
Parameter Value Parameter Model
Carrier frequency fc 780 MHz LLOS max (23.9 − 1.8 log10 (hUE ), 20) log10 (d)
gNB +20 log10 ( 40πf
3
c
)
Transmit power PT,gNB 46 dBm LNLOS max(LLOS , −12 + (35 − 5.3 log10 (hUE ))
Mast height hgNB 35 m log10 (d) + 20 log10 ( 40πf
3
c
))
Sector antenna directions 30◦ , 150◦ , 270◦ in azimuth angles σX,LOS 4.2 exp(−0.0046hUE )
Mechanical down-tilt 3◦ σX,NLOS 6.0
Sector antenna model 8×1 ULA with element spacing of 0.5λ
aUE
Height hUE variable
bility is given as:
√ √
γFW − γmin
 
Antenna model Ne × Ne uniform rectangular array with 1
pout = √erfc , (5)
variable size and 0.5λ element spacing 2 2σSIR
where erfc() is the complementary
q error function, γFW =
FW S 2 FW 2
where PT,dBm is the gNB transmit power in dBm, GT,dBi P̄S,dBm −P̄I,dBm and σSIR = (σX ) + (σX ) are the mean
and GR,dBi are the antenna gains at the gNB and the aUE, SIR and its standard deviation, respectively, according to the
respectively, in dBi while LdB is the path loss in dB. Fenton-Wilkinson approximation.
1) Link Quality: In interference-dominant environments, 3) Availability: The availability is defined as the ratio,
the SIR can be used for rating the link quality. The average in percentage, of the time period in which the end-to-end
SIR in dB is given by communication service is delivered with an agreed Quality of
Service (QoS) to the total period of time in which the system
γ = P̄S,dBm − P̄I,dBm , (2)
is expected to deliver the end-to-end service [23]. In this study,
where P̄S =10P̄S,dBm /10 is the average power of the serving cell the availability is determined for the downlink quality by using
signal in mW and P̄I =10P̄I,dBm /10 is the average interference the outage probability as measure of agreed QoS with respect
power in mW. to the minimum SIR threshold γmin . Hence, the availability is
Let NgNB be the cardinality of the set G=GSw ∪GFo . Then, given by the duration for which the pout is below a threshold
the average power received from any gNB sector can be de- pth
out , divided by the total duration of the approach trajectory.
noted as P̄Rg,s , where g∈G= {1, . . . , NgNB } identifies the gNB
and s ∈ S = {1, 2, 3} identifies its sector. The interference D. Channel
power P̄I in (2) is computed as the sum of received powers
The ATG channel model used for evaluating the link budget
from all sectors in the network except the sectors of the serving
is altitude dependent and constructed from existing 3rd Gen-
gNB. If ′the′ gNB g ′ serves the aUE via its sector s′ , then
eration Partnership Project (3GPP) channel models as follows:
P̄S = P̄Rg ,s and
X X g̃,s̃ For aUE altitudes hUE ≤ 300 m, the channel model of a rural
P̄I = P̄R . (3) macro scenario for UAVs is adopted from [2]. The model from
g̃∈G\g ′ s̃∈s [12] is used for hUE > 300 m. The threshold altitude is chosen
2) Outage Probability: The outage probability pout is de- because the UAV channel model is valid for hUE ≤ 300 m [2].
fined as the probability of the instantaneous SIR falling below A probabilistic path loss is used as in [24];
a threshold value γmin . The instantaneous SIR can be obtained LdB = pLOS LLOS + (1 − pLOS )LNLOS , (6)
g,s
from the instantaneous received power PR,dBm given by:
where LLOS and LNLOS are the path loss values in Line of Sight
g,s g,s g,s
PR,dBm = P̄R,dBm + XdB , (4) (LOS) and Non Line of Sight (NLOS) conditions, respectively,
g,s
where P̄R,dBm is computed according to (1) and XdB g,s
is a while pLOS denotes the probability of LOS conditions. A
random variable distributed according to N (0, σX ) that cap- g,s similar approach is used for computing the average shadow
tures shadow fading. The notation N (µ, σ) denotes a Gaussian fading standard deviation σX :
distribution with mean µ in dB and standard deviation σ in dB. σX = pLOS σX,LOS + (1 − pLOS )σX,NLOS , (7)
Thus, the received power level of the serving gNB  and of each
g,s g,s
 where σX,LOS and σX,NLOS are the standard deviations of the
interfering gNB are distributed according to N P̄R,dBm , σX, . shadow fading in LOS and NLOS conditions, respectively.
The probability density function of the total instantaneous The path loss and shadow fading models adopted from [2]
interference power PI,dBm is obtained using the Fenton- for hUE ≤ 300 m are given in Table II. The values of pLOS are
Wilkinson approximation [22] so that it is distributed accord- taken from Table B-1 of [2] which saturate to 1 at hUE =40 m.
FW FW FW FW
ing to N (P̄I,dBm , σX ). Here, P̄I,dBm and σX are calculated For hUE >300 m, free space propagation in LOS conditions is
g̃,s̃ g̃,s̃
from PR and σX with g̃∈G \ g , s̃∈S using the Fenton-
′
assumed [12]. The standard deviation σX,LOS is obtained by
S g ′ ,s′
Wilkinson method. Finally, with σX =σX the outage proba- linear interpolation from Table III.
 TABLE III 20 γFW γ

SIR in dB
σX,LOS FOR hUE > 300 m AND ELEVATION ANGLES θ
θ 10◦ 20◦ 30◦ 40◦ 50◦ 60◦ 70◦ 80◦ 90◦ 10
σX,LOS 1.79 1.14 1.14 0.92 1.42 0.56 0.85 0.72 0.72
0
E. Handovers 0 20 40 60 80 100 120 140 160 180 200
Time in s
As the aUE moves over the TMCN, the serving gNB
Fig. 3. SIR with a 8×8 AESA along the Rome trajectory.
may need to be changed via handover procedure to ensure
a sufficient link quality. In this work, the handover procedure 30
Omni 2×2 4×4 8×8 16 × 16
is modeled as follows:

SIR γ in dB
First: At any point on the trajectory, the receive signal power 15
levels of all gNBs in radio range are obtained from one antenna
element of the AESA. Thus, the P̄Rg̃,s̃ values are subject to the 0
directivity of the antenna elements only. −15
Second: From the set of gNBs whose power levels were
obtained, a subset GH containing |GH |=64 gNBs with the 0 100 200 300 400 500 600 700 800
strongest power levels is created. For each gNB in GH , the P̄Rg̃,s̃ Time in s
values are calculated assuming the AESA to be configured Fig. 4. SIR along the Rome trajectory for various antenna arrays at the aUE
such that its main beam is directed towards the gNB. Using
these values, the SIR is computed with the gNB signal as C. Scheduling Support
wanted component and the sum of the signals of all other gNBs
In the scheduling-support scenario, the network supports
as interference component. The gNB in GH with the highest
the aUEs in terms of dedicated time-frequency resources.
SIR is selected as a handover candidate for the particular
The resources allocated to the aUE are orthogonal to those
trajectory point. The rather large set size was chosen to make
allocated to the terrestrial UEs. Thus, traffic generated by
sure that the cells offering high SIR remain in the set GH .
terrestrial UEs does not cause interference at the aUE.
Third: A handover is performed, if the SIR from the The allocation of dedicated resources can be realized in
handover candidate exceeds the SIR from the serving gNB by the form of a network slice. In our model, a network slice is
more than 3 dB. The procedure is constrained by a minimum assumed to be provided by a set of gNBs GSlice
r
⊆ GSw . The set
time interval of 1 s between subsequent handovers. GSlice contains all the gNBs that fall within r=50 km of the
r

Zurich airport. In this scenario, the serving gNB can only be

III. E VALUATION
among the gNBs hosting the slice, hence g ′ ∈ GSlice
r
. Hence, an
The ATG link quality is evaluated on three flight trajectories, interfering gNB will belong to the set g̃ ∈ (GSw \ GSlice
r
) ∪ GFo .
where aircraft approach one of the three runways of the Zurich
IV. R ESULTS AND D ISCUSSION
airport. The trajectories are based on real data obtained from
Flightradar24.com for flights departing from Rome, Venice Results are presented for the ATG link quality in terms of
and Philadelphia. The approach segments of these trajectories average SIR and availability, and handover behavior observed
are depicted in Fig. 2. The evaluation considers scenarios with for the three trajectories. Fig. 3 depicts an example of the
different levels of support provided by the TMCN. Recall that average SIR γ according to (2) as well as its approximation
g ′ and g̃ denote the serving and interfering gNB, respectively, γFW . The figure also includes the ±2σSIR envelope (shaded
while G=GSw ∪GFo . The scenarios are as follows: regions) for γFW . Fig. 3 shows that depending upon the σSIR
the average SIR needs to be sufficiently higher than γmin for
A. No Support avoiding communications outages. Further, it can be seen that
γFW underestimates the computed SIR γ only slightly.
The no-support scenario is the baseline case where the aUE
is treated as normal terrestrial UE. Hence, g ′ ∈GSw and g̃∈G\g ′ . A. Link quality
The link quality is evaluated with respect to the AESA
B. Roaming Support configurations listed in Table IV and level of support by the
In the roaming-support scenario, the foreign network is TMCN in the following.
not merely a source of interference but can also provide a 1) No Support: Fig. 4 shows the average SIR along the
serving gNB to the aUE. In this case, g ′ ∈G while g̃∈G\g ′ . Rome to Zurich trajectory for the different AESA configura-
Roaming support can be beneficial at airports which are close tions.The figure also includes the SIR using an omnidirectional
to international borders such as the Zurich airport. It is ac- antenna at the aUE as a reference. The SIR observed in
knowledged that cross border handovers are more complicated omnidirectional setup highlights the severity of interference
than intra-operator handovers [25]. However, the two classes at high altitudes. However, the AESA improves the SIR
of handovers are performed identically in this work. significantly where arrays with Ne =4, 16, 64 and 256 elements
 TABLE IV 100

Availability in %
A IRCRAFT AESA CONFIGURATION WITH λ = 0.38 m
Array configuration 2×2 4×4 8×8 16 × 16 85
No. of Elements Ne 4 16 64 256
Physical dimension in m2 (0.19)2 (0.58)2 (1.34)2 (2.88)2
70
pth
out =10
−3
pth
out =10
−5
pth
out =10
−7

55
No support 50km slice Roaming −6 −5 −4 −3 −2 −1 0 1 2
30 γmin in dB
SIR γ in dB

20 Fig. 8. Availability with different values for γmin and pth

out for the Venice
10 trajectory using 8 × 8 AESA in no support scenario.
TABLE V
0 AVAILABILITY IN % FOR pth −5 W. R . T. γ
min = 0 ( AND −6 dB)
out = 10
−10 Scenario Rome Venice Philadelphia
0 100 200 300 400 500 600 700 800 No support 55.5 (78.6) 73.5 (93.5) 81.3 (97.5)
Time in s Roaming 59.9 (86.1) 73.5 (93.5) 81.3 (97.5)
Fig. 5. SIR for the Rome trajectory with a 8×8 AESA. Scheduling 71.7 (91.7) 92.1 (100) 97.7 (100)

increase the SIR by 14.1, 20.2, 27.6 and 35.2 dB in average as

compared to the omnidirectional antenna. to 30 dB. This improvement is, however, limited to the portion
The SIR curves in Fig. 4 exhibit a visible dip between of the trajectory near the border. Support by network slicing
the 500 and 700 s marks that is independent from the AESA increases the SIR on average by 6.5 dB along the complete
configuration. This is due to the fact that this portion of the trajectory with up to 14 dB improvement at the 337 s mark.
trajectory is close to and even partially across the Swiss- Fig. 6 shows an average improvement of 6.29 dB from the
German border where the Swiss network has poor coverage. scheduling support for the Venice trajectory with up to 16 dB
The physical dimension along each side of the AESA increase at the 436 s mark. The roaming support does not affect
configurations in Table IV is estimated using the relation the SIR since the trajectory is far away from the border.
√ For the Philadelphia trajectory, Fig. 7 shows that the
0.5λ( Ne − 1). Assuming that the physical size of the 16×16
array is too large to be feasible, the 8×8 antenna array is scheduling support improves the SIR by up to 14.4 dB (at 124 s
chosen for the rest of evaluations. The average SIR observed mark) with an average SIR improvement of approx. 7 dB. Only
along the Rome, Venice and Philadelphia is shown in Figs. 5, marginal improvements of the SIR can be observed in case of
6 and 7, respectively, with respect to the different levels support by roaming.
of support by the TMCN. Without support by the network, 3) Availability: Fig. 8 shows the link availability for mul-
negative SIR values can be observed in each trajectory. tiple pth
out w.r.t. a range of γmin . It can be seen that stringent

2) Roaming & Scheduling Support: For the Rome trajec- QoS requirements, i.e., low pth out w.r.t. a high γmin , lead to low

tory, Fig. 5 shows that roaming support improves the SIR by up availability and vice versa. Table V presents the availability
of the ATG link if pth out =10
−5
is assumed together with a
rather strict minimum SIR requirement of γmin =0 dB. For
No support 50km slice Roaming reference, the table also includes availability computed with
30 γmin =−6 dB [2]. Without support by the TMCN, the aerial
SIR γ in dB

link is available 55.5, 73.5 and 81.3 percent of the time for
20 the Rome, Venice and Philadelphia trajectory, respectively.
10 Roaming support slightly increases the availability only for the
Rome trajectory. Support by scheduling significantly increases
0 the link availability to 71.7, 92.1 and 97.7 percent of the time
0 100 200 300 400 500 600 700 800 900 for the Rome, Venice and Philadelphia trajectory, respectively.
Time in s
B. Handovers
Fig. 6. SIR for the Venice trajectory with a 8×8 AESA.
Table VI contains handover data for the flight trajectories
No support 50km slice Roaming in the three evaluation scenarios. The table provides the mean
30
SIR γ in dB

of the time intervals between handovers T̄HO and the largest

20 interval between consecutive handovers Tmax in seconds. The
smallest inter-handover interval across all the scenarios was
10 observed to be 1 s.
0 For each scenario, the corresponding row in Table VI shows
that the handover frequency increases from left to right. For
0 100 200 300 400 500 600 700 800
instance, the trajectories from Rome, Venice and Philadelphia
Time in s
require on average a handover every 12.23, 11.42 and 8.65 s,
Fig. 7. SIR for the Philadelphia trajectory with a 8×8 AESA.
 TABLE VI [3] I. Kovacs, R. Amorim, H. C. Nguyen, J. Wigard, and P. Mogensen, “In-
H ANDOVER INTERVALS IN SECONDS IN VARIOUS SCENARIOS terference Analysis for UAV Connectivity over LTE Using Aerial Radio
Measurements,” in 2017 IEEE 86th Vehicular Technology Conference
Rome Venice Philadelphia
(VTC-Fall), 2017, pp. 1–6.
Scenario T̄HO Tmax T̄HO Tmax T̄HO Tmax [4] X. Lin, V. Yajnanarayana, S. D. Muruganathan, S. Gao, H. Asplund,
No support 12.23 67 11.42 45 8.65 37 H.-L. Maattanen, M. Bergstrom, S. Euler, and Y.-P. E. Wang, “The
Sky Is Not the Limit: LTE for Unmanned Aerial Vehicles,” IEEE
Roaming 11.53 67 11.42 45 8.56 37
Communications Magazine, vol. 56, no. 4, pp. 204–210, 2018.
Scheduling 11.4 44 10.42 42 7.82 35 [5] S. D. Muruganathan, X. Lin, H.-L. Määttänen, J. Sedin, Z. Zou, W. A.
Hapsari, and S. Yasukawa, “An Overview of 3GPP Release-15 Study on
Enhanced LTE Support for Connected Drones,” IEEE Communications
respectively, in the baseline scenario. Together with Table V, Standards Magazine, vol. 5, no. 4, pp. 140–146, 2021.
[6] Qualcomm, “LTE Unmanned Aircraft Systems,” Qualcomm Technolo-
the handover data reveals that high availability comes with gies, Inc, Trail Report 38.811, 5 2017, version 1.0.1.
more frequent handovers. The same trend is seen by consider- [7] Y. Zeng, Q. Wu, and R. Zhang, “Accessing From the Sky: A Tutorial on
ing columns of the same trajectory in the two tables. Recall (cf. UAV Communications for 5G and Beyond,” Proceedings of the IEEE,
vol. 107, no. 12, pp. 2327–2375, 2019.
Sec II-E) that handovers are executed whenever a candidate [8] R. Muzaffar, C. Raffelsberger, A. Fakhreddine, J. L. Luque, D. Emini,
gNB is better by 3 dB from the serving one regardless of the and C. Bettstetter, “First Experiments with a 5G-Connected Drone,”
case that the SIR from the serving gNB might be above the in 6th ACM Workshop on Micro Aerial Vehicle Networks, Systems,
and Applications, ser. DroNet ’20. New York, USA: Association for
γmin . This, combined with the aUE traversing the sidelobes Computing Machinery, 2020.
of the gNB’s antenna gain pattern leads to the direct relation [9] V. Marojevic, I. Guvenc, R. Dutta, M. L. Sichitiu, and B. A. Floyd,
between the availability and handover frequency. “Advanced Wireless for Unmanned Aerial Systems: 5G Standardization,
Research Challenges, and AERPAW Architecture,” IEEE Vehicular
V. C ONCLUSION Technology Magazine, vol. 15, no. 2, pp. 22–30, 2020.
[10] 3GPP, “5G; Unmanned Aerial System (UAS) Support in 3GPP,” 3rd
This work evaluated the feasibility of using ATG commu- Generation Partnership Project, TS 22.125, 04 2022, ver. 17.6.0.
[11] “European aviation network,” https://www.europeanaviationnetwork.com,
nications to support the safe landing of a passenger aircraft in accessed: 2023-04-28.
approach to an airport by means of a remote operator. The fo- [12] 3GPP, “Study on Air-to-ground network for NR,” 3rd Generation Part-
cus was on the interference affected downlink established with nership Project, TR 38.876, 01 2022, ver. 16.16.0.
[13] V. Yajnanarayana, Y.-P. Eric Wang, S. Gao, S. Muruganathan, and
a 5G network that is optimized for terrestrial coverage. The X. Lin Ericsson, “Interference Mitigation Methods for Unmanned Aerial
results were obtained using link budget simulations yielding Vehicles Served by Cellular Networks,” in 2018 IEEE 5G World Forum
SIR and availability values for real approach trajectories. It is (5GWF), 2018, pp. 118–122.
[14] H. C. Nguyen, R. Amorim, J. Wigard, I. Z. KováCs, T. B. Sørensen,
shown that an acceptable level in SIR can be achieved with and P. E. Mogensen, “How to Ensure Reliable Connectivity for Aerial
an 8 × 8 AESA installed onboard the aircraft, providing up to Vehicles Over Cellular Networks,” IEEE Access, vol. 6, pp. 12 304–
81.3% availability. 12 317, 2018.
[15] T. Izydorczyk, M. Bucur, F. M. L. Tavares, G. Berardinelli, and
The presented results demonstrate that significant improve- P. Mogensen, “Experimental Evaluation of Multi-Antenna Receivers for
ments in SIR and availability can be achieved when the UAV Communication in Live LTE Networks,” in 2018 IEEE Globecom
TMCNs supports ATG communications by provisioning ded- Workshops (GC Wkshps), 2018, pp. 1–6.
[16] Y. Huang, Q. Wu, T. Wang, G. Zhou, and R. Zhang, “3D Beam Tracking
icated resources, e.g., in form of a network slice. With the for Cellular-Connected UAV,” IEEE Wireless Communications Letters,
dedicated resources, a downlink availability of up to 97.7% vol. 9, no. 5, pp. 736–740, 2020.
was achieved. Depending upon proximity of international [17] T. Izydorczyk, G. Berardinelli, P. Mogensen, M. M. Ginard, J. Wigard,
and I. Z. Kovács, “Achieving High UAV Uplink Throughput by Using
borders, roaming support might be crucial. Further, this work Beamforming on Board,” IEEE Access, vol. 8, pp. 82 528–82 538, 2020.
indicated a direct relation between handover frequency and [18] A. Colpaert, E. Vinogradov, and S. Pollin, “3D beamforming and han-
link availability. dover analysis for UAV networks,” in 2020 IEEE Globecom Workshops,
2020, pp. 1–6.
Finally, it is pointed out that this work assumed the gNBs [19] Swiss Federal Office of Communications, “Location of radio
to be equipped with down-tilted antennas. Measures to sup- transmitters,” accessed 30 March 2022. [Online]. Avail-
port aerial users in TMCN, such as up-tilted antennas and able: https://www.bakom.admin.ch/bakom/en/homepage/frequencies-
and-antennas/location-of-radio-transmitters.html
electrical beam steering at the gNBs together with favorable [20] 3GPP, “5G; Study on channel model for frequencies from 0.5 to 100
gNB locations are expected to drastically improve link quality GHz,” 3rd Generation Partnership Project, TR 38.901, 11 2020, ver.
and reduce the handover frequency in ATG communications. 16.1.0.
[21] ITU-R, “Characteristics of terrestrial IMT-Advanced systems for fre-
Moreover, carrier frequencies higher than 780 MHz, as con- quency sharing/interference analyses,” International Telecommunication
sidered in [12], can enable the use of AESAs larger than Union, Tech. Rep. M.2292-0, Dec. 2013.
8 × 8 which can also be helpful. With these measures, a link [22] G. L. Stüber, Principles of Mobile Communication, 4th ed. Springer-
Verlag GmbH, 2017.
availability approaching 100% can be expected. [23] 3GPP, “5G; Service requirements for the 5G system,” 3rd Generation
Partnership Project, TS 22.261, 01 2022, ver. 16.16.0.
R EFERENCES [24] W. Shin and M. Vaezi, “UAV-Enabled Cellular Networks,” in 5G and
[1] RTCA, Minimum Aviation System Performance Standards for C2 Link Beyond. Springer, 2021, pp. 165–200.
Systems Supporting Operations of Unmanned Aircraft Systems in U.S. [25] A. Kousaridas, M. Fallgren, E. Fischer, F. Moscatelli, R. Vilalta,
Airspace, Radio Technical Commission for Aeronautics Std. DO-377, M. Mühleisen, S. Barmpounakis, X. Vilajosana, S. Euler, B. Tossou, and
Mar. 2019. J. Alonso-Zarate, “5G Vehicle-to-Everything Services in Cross-Border
[2] 3GPP, “Enhanced LTE support for aerial vehicles,” 3rd Generation Environments: Standardization and Challenges,” IEEE Communications
Partnership Project, TR 36.777, 01 2018, ver. 15.0.0. Standards Magazine, vol. 5, no. 1, pp. 22–30, 2021.