Integrating Terrestrial and Non-terrestrial Networks:

3D Opportunities and Challenges

Abstract—Integrating terrestrial and non-terrestrial networks enhancements towards a fully integrated terrestrial plus non-
has the potential of connecting the unconnected and enhancing terrestrial network (NTN) able to satisfy both ground and
the user experience for the already-connected, with technological aerial requirements. At first glance, terrestrial networks (TNs)
and societal implications of the greatest long-term significance. A
convergence of ground, air, and space wireless communications could be: (i) re-engineered and optimized to support aerial
also represents a formidable endeavor for the mobile and satellite users [5], [6], or (ii) complemented by NTN infrastructure
communications industries alike, as it entails defining and intel- such as low Earth orbit (LEO) satellite constellations or aerial
ligently orchestrating a new 3D wireless network architecture. base stations (BSs) to further enhance performance [7], [8].
In this article, we present the key opportunities and challenges Cost-related factors may advocate for a progressive roadmap.
arising from this (r)evolution by presenting some of its disruptive
use-cases and key building blocks, reviewing the relevant stan- In the present paper, we discuss the opportunities and
dardization activities, and pointing to open research problems. By challenges lying behind a 3D integrated TN-NTN. We begin
arXiv:2207.10385v1 [cs.IT] 21 Jul 2022

considering two multi-operator paradigms, we also showcase how by providing examples of key use-cases, overviewing the
terrestrial networks could be efficiently re-engineered to cater building blocks of an integrated TN-NTN architecture, and
for aerial services, or opportunistically complemented by non- summarizing the most relevant 3GPP standardization activ-
terrestrial infrastructure to augment their current capabilities.
ities. We then introduce the case study of a conventional
terrestrial operator pursuing aerial connectivity through two
I. I NTRODUCTION plausible choices: (i) deploying dedicated uptilted cells—or
A mobile connection is our window to the world. The partnering with a specialized aerial operator doing so—reusing
current social, economic, and political drive to reach global the same spectrum; (ii) leasing infrastructure or solutions from
wireless coverage and digital inclusion acknowledges con- a LEO satellite operator. We conclude by reviewing the main
nectivity as vital for accessing fair education, medical care, hurdles that stand in the way to an integrated TN-NTN and
and business opportunities in a post-pandemic society. Sadly, pointing out key open problems worthy of further research.
nearly half of the population on Earth remains unconnected.
Indeed, rolling out optical fibers and radio transmitters to II. U SE - CASES , A RCHITECTURE , AND S TANDARDIZATION
every location on the planet is not economically viable, and In this section, we describe the main use-cases and compo-
reaching the billions who live in rural or less privileged nents of a plausible integrated TN-NTN, and we summarize
areas has remained a chimera for decades. The long-overdue the major NTN and UAV standardization advancements.
democratization of wireless communications requires a wholly
new design paradigm to realize ubiquitous and sustained A. Use-cases
connectivity in an affordable manner. The opportunities unlocked by integrating TN and NTN
Meanwhile, in more urbanized and populated areas, even capabilities could lead to a vast number of new applications
5G may eventually fall short of satiating our appetite for and services. In what follows, we provide a representative
mobile internet and new user experiences. Life in the 2030s down-selection of the key use-cases.
and beyond will look quite different from today’s: hordes Critical communications: Connectivity from space or air can
of network-connected UAVs (uncrewed aerial vehicles) will empower ultra-reliable critical communications in the absence
navigate 3D aerial highways—be it for public safety or to of cellular coverage or during an emergency or natural disaster.
deliver groceries to our doorstep—, and flying taxis will re- In this case, when the ground network becomes dysfunctional
shape how we commute and, in turn, where we live and and the importance of providing rapid and resilient connec-
work. The bold ambition of reaching for the sky will take tivity cannot be overstated, NTNs can ensure replacement
the data transfer capacity, latency, and reliability needs for the coverage through direct access from space/air or even via
underpinning network to an extreme, requiring dedicated radio satellite- or cellular- backhauled UAV radio access nodes.
resources and infrastructure for aerial services [1], [2]. Massive IoT and immersive communications: NTNs can
In a quest for anything, anytime, anywhere connectivity— cover large areas of land or sea populated with both static
even up in the air—, next-generation mobile networks may and nomadic sensory nodes, all collecting real-time data.
need to break the boundary of the current ground-focused Aggregating and displaying the latter through AR/VR applica-
paradigm and fully embrace aerial and spaceborne commu- tions will provide users with spatial and contextual awareness,
nications [3], [4]. To this end, the wireless community has enabling immersive human-machine interaction, likely one of
already rolled up its sleeves in (re)search for technology the 6G killer apps. Depending on the latency requirements and
 • Geostationary orbit (GSO) satellites, orbiting the equato-
rial plane at an altitude of about 35786 km, and creating
fixed beams with a footprint radius of up to 3000 km.
• Non-GSO satellites, such as LEO, deployed at altitudes
between 300–1500 km, creating footprints of up to 1000
km radius per beam. Unlike their GSO counterpart, LEO
satellites move fast with respect to a given point on the
Earth, with an orbital period of just a few hours, and thus
require large constellations for coverage continuity.
• Aerial BSs such as HAPSs, placed in the stratosphere at
around 20 km, and creating multiple cells sized about 10
km each, or UAV radio access nodes, flying at heights
somewhere between 0-1 km.
• Ground gateways connecting aerial and spaceborne plat-
forms to the core network through so-called feeder links.
Terminals: The end-devices of a 3D TN-NTN can be
classified as follows:
• Stationary and vehicular ground users (GUEs), in areas

Fig. 1: Exemplified integrated TN-NTN. NTN BS functionalities can ranging from dense urban to suburban, rural, and remote.
be placed onboard satellites or at the NTN gateway, respectively • UAVs, eVTOLs, and aircrafts, demanding in-flight con-
entailing a regenerative or transparent satellite payload [7]. nectivity at altitudes of few hundred meters, 1–3 km, and
10–12 km, respectively [5].
Satellite-connected devices can either be handheld/IoT or
sensory node capabilities, data aggregation could be handled
equipped with a very-small-aperture terminal (VSAT), de-
by a LEO constellation in the field of view of a ground
pending on the use-case and the carrier frequency of the
gateway or aerial BSs. NTN broadcast/multicast could then
service link. Indeed, the more benign link budget in the S-
pursue content scalability and uninterrupted delivery to users
band (sub-6 GHz) enables direct access to omni- or semi-
in cars, trains, and vessels.
directional handheld/IoT terminals. Operating in the Ka-band
Aerial communications: Beyond standalone TNs, primarily
(mmWave spectrum) incurs a higher attenuation, which must
designed for 2D usage, an integrated TN-NTN could support
be compensated with a larger antenna gain by employing
reliable data and control links to multiple UAVs, electrical
a VSAT. The latter can be either fixed or mounted on a
vertical take-off and landing vehicles (eVTOLs), and aircrafts.
moving platform (buses, trains, vessels, or aircrafts), thus
These services would be guaranteed in specific 3D areas—
giving options for either mobile or fixed broadband access.
aerial corridors or waypoint trajectories—where end-devices
will be allowed to fly at different heights. The potential C. Standardization
of UAVs may only truly be unleashed once the network
capabilities and regulations allow for autonomous operation Standardization work on non-terrestrial communications in
beyond visual line-of-sight (LoS) [9], [10]. 3GPP dates back to 2017 [11]. This effort can be classified
nowadays into two main areas, namely NTN enhancements
B. Architecture and TN support for UAVs. The former aims at defining a
A simplified integrated TN-NTN architecture is illustrated global standard for future spaceborne communications, fos-
in Fig. 1, with service links connecting a user terminal— tering an explosive growth in the satellite industry. Activities
either handheld/IoT or VSAT—to TN/NTN BSs, feeder links within the latter serve the twofold purpose of ensuring that
connecting the NTN segment to the ground core network, and mobile standards meet the connectivity needs for safe UAV
(optionally) inter-satellite and/or inter-high-altitude platform operations, and that other users of the network do not expe-
stations (HAPS) links. rience a loss of service due to their proximity to UAVs. The
Network platforms: The 3D TN-NTN will avail of a multi- objectives and outputs of the 3GPP work carried out from Rel-
layered multi-band infrastructure, arranged hierarchically, with 15 up to Rel-17, along with the topics currently under study
the following nodes operating at different altitudes and offer- for Rel-18 are outlined as follows and summarized in Table I.
ing user-centric coverage and service:
• TN BSs of various size, power, height, and orienta- NTN enhancements: In 3GPP parlance, the term NTN refers
tion, operating in sub-6 GHz, mmWave, and eventually to utilizing satellites or HAPS to offer connectivity services
THz bands, and deployed with different densities. Along and complement terrestrial networks, especially in remote
with conventional downtilted BSs, mobile operators may areas where cellular coverage is unavailable. In Rel-17, 3GPP
choose to deploy dedicated infrastructure, e.g., uptilted introduced a set of basic features to enable 5G NR operation
cells, to serve aerial users. over NTNs in FR1, i.e., up to 7.125 GHz. 3GPP Rel-18
 TABLE I: Overview of the 3GPP standardization work on NTN and UAV communications [1], [11], [12].

Release Non-terrestrial Networks (NTN) Enhancements Support for Uncrewed Aerial Vehicles (UAVs)
Study on New Radio (NR) to support NTNs [TR 38.811] Enhanced LTE support for aerial vehicles [TR 36.777]
Identified relevant scenarios for NTN deployment studies and in- Proposed user- and network-based solutions for downlink and uplink
tegration in terms of: frequency bands (S-band at 2 GHz vs. Ka- interference mitigation, mobility, and UAV identification. Specified
band at 10–20 GHz), typical footprint sizes and minimum elevation channel models based on TR 38.901 accounting for the UAV height.
Rel-15
angles, antenna models and beam configurations (Earth-fixed, steered Enhancements to measurement report triggering [TS 36.331]
towards a fixed area on the ground, vs. moving beams), and NTN ter- Enhancements included the addition of two reporting events—H1
minals (handheld vs. VSAT). Specified propagation channel models (above) and H2 (below) user height thresholds—to help the network
based on TR 38.901 with NTN-specific modifications. identify a UAV and deal with any potential interference.
Solutions for NR to support NTNs [TR 38.821] Remote identification of UAVs [TS 22.825]
Focused on FR1 for handheld and IoT satellite access. Identified Studied the potential requirements and use-cases for remote UAV
modifications required for the physical and higher layers and system- identification and the services to be offered accordingly. Aimed at
level simulation assumptions. Studied the impact of delay on random allowing air traffic control and public safety agencies to query the
access, scheduling, and hybrid automatic repeat request (HARQ), as identity and metadata of a UAV and its controller via the UTM to
Rel-16
well as mobility management for moving LEO platforms. authorize, enforce, and regulate UAV operation.
Using satellite access in 5G [TR 22.822] UAV connectivity, identification, and tracking [TR 23.754]
Identified use-cases for the provision of services when considering Considered 3GPP-supported connectivity between UAVs and the
the integration of 5G satellite-based access components. Identified UAV traffic management (UTM), as well as the detection and
new services and the corresponding requirements. reporting of unauthorized UAVs to the UTM.
NB-IoT and eMTC support for NTN [TR 36.763] 5G enhancements for UAVs [TS 22.125, TS 22.829]
Focused on IoT applications by addressing issues related to LTE Produced new UAV KPIs and communication needs related to:
timing relationships, uplink synchronization, and HARQ. payload, command and control traffic, on-board radio access node
Architecture aspects for using satellite access in 5G [TR 23.737] (UxNB), and service restrictions and network exposure for the UAV.
Rel-17
Specified enhancements for RF and physical layer, protocols, radio Application layer support for UAVs [TR 23.755]
resource management, and frequency bands. Identified a suitable Studied use-cases for UAV identification and tracking, their impact
architecture, addressed TN-NTN roaming and timing-related issues, on the application layer, and UAV-UTM interactions for route autho-
enhanced conditional handover, and location-based triggering. rization, location management, and group communication support.
NR NTN enhancements NR support for UAVs
Will study NR NTN coverage for realistic handheld terminals and Will study enhancements on measurement reports, signaling for
Rel-18 access above 10 GHz to fixed and moving platforms. Will investigate subscription-based UAV identification and its multicast, additional
requirements for network-verified user location, and address mobility triggers for conditional handover, and beam management at FR1,
and service continuity between TN-NTN and across different NTNs. including UAV directional antennas and BS uptilt beamforming.

will enhance 5G NR NTN operation by improving coverage The main system-level assumptions for these two setups are
for handheld terminals, studying deployments above 10 GHz, summarized in Table II.
addressing mobility and service continuity between TN-NTN
A. Example I: Re-designing TNs for NTN Terminals
as well as across different NTNs, and investigating regulatory
requirements for network-verified user location [12]. As the penetration of aerial users increases, a terrestrial
Support for UAVs: 3GPP introduced 4G LTE support for mobile network operator (MNO) may choose to cater for UAV
UAVs back in Rel-15, including signaling for subscription- connectivity or partner with another MNO intending to do so
based aerial user identification, reporting of UAV height, [5], [13]. The latter gives rise to the following hypothetical
location, speed, and flight path, and new measurement reports setup with two operators sharing the same spectrum, namely:
to address aerial interference up to a certain density of low- • A terrestrial operator, MNOT , running a standard network
altitude UAVs. In subsequent releases, 3GPP addressed appli- comprised of downtilted cells to serve legacy GUEs.
cation layer support and security for connected UAVs, also • An aerial operator, MNOA , running a dedicated network
defining the service interactions between UAVs and the UAV of uptilted BSs reserved exclusively for connected UAVs.
traffic management (UTM) system. As 5G use-cases evolve, The deployment sites of both operators are on a hexagonal
Rel-18 will introduce 5G NR support for devices onboard layout and comprised of three co-located BSs, each covering
aerial vehicles, studying additional triggers for conditional one sector (i.e., a cell) spanning an angular interval of 120◦ .
handover, BS uptilting, and signaling to indicate UAV beam- Let ISDT and ISDA denote the respective inter-site distances,
forming capabilities, among other enhancements [12]. whereby we fix the former to 500 m, and vary the latter
to study its effect. We assume 15 GUEs for each MNOT
III. O PPORTUNITIES cell, and for all values of ISDA , we keep the UAV density
constant and according to 3GPP Case 3 in TR 36.777, yielding
In this section, we consider two multi-operator case stud- {1, 4, 9} UAVs/cell under ISDA = {500, 1000, 1500} m,
ies to illustrate how terrestrial networks could be (i) effi- respectively. GUEs are located both outdoor at 1.5 m and
ciently re-engineered to support non-terrestrial end-devices indoor in buildings consisting of several floors. UAVs fly
such as UAVs, or (ii) opportunistically complemented by non- outdoor at a height of 150 m. We assume all GUEs and UAVs
terrestrial infrastructure to augment their current capabilities. to have a single omnidirectional antenna, and to connect to
TABLE II: System-level parameters. All main assumptions follow
3GPP Technical Reports 38.901, 36.777, 38.811, and 38.821.
networking decisions are performed individually by each
BS. On the other hand, such a simplification comes at the
MNOT & MNOA cost of inter-MNO co-channel interference.
Cell layout Hexagonal, 3 sectors/site, 1 BS/sector at 25 m • Eigendirection-aware (EDA) precoding, where BSs ded-
Intersite distance ISDT = 500 m, ISDA = {1500,1000,500} m icate a certain number of spatial degrees of freedom to
Frequency band 100 MHz TDD at 3.5 GHz place radiation nulls, thereby canceling interference on
Spectrum MNOT and MNOA in the same band the dominant eigendirections of the inter-cell channel
Scheduler DL: 50 MHz per GUE, 50 MHz per UAV subspace [14]. This approach requires coordination be-
(round robin) UL: 10 MHz per GUE, 50 MHz per UAV
tween MNOT and MNOA for channel state information
DL: ZF (8 users) or EDA (8 users + 16 nulls)
Precoding acquisition, possibly entailing them to belong to the same
UL: ZF (4 users) or EDA (4 users + 8 nulls)
Downlink power MNOT : 46 dBm, MNOA : 46 dBm or less
network provider.
Uplink power Fractional power control with α = 0.80, P0 = We focus our analysis on the uplink, the more data-hungry
−100 dBm, and Pmax = 23 dBm direction for UAVs, whose generated transmissions may pose
Antenna elements Horiz./vert. HPBW: 65◦ , max gain: 8 dBi
a threat to legacy GUEs [1]. Fig. 2 shows the SINR attained
Antenna array 8 × 8 X-POL, fully digital
by UAVs and GUEs for various values of ISDA and the two
Antenna tilt MNOT : 12◦ (down), MNOA : −45◦ (up)
precoding schemes. These results show the following:
Noise figure 7 dB
MNOS • Offloading UAVs from MNOT sees their SINR reduced,
Cell layout Orbit: 600 km, 7 beams centered on a hexag- unless the deployment of MNOA is sufficiently dense. Im-
onal grid, elevation angle: variable portantly, UAVs remain in coverage even under ISDA =
FRF=1: 30+30 MHz (DL+UL) FDD at 2 GHz 1500 m. Offloading, however, provides UAVs with higher
Frequency band
FRF=3: 10+10 MHz (DL+UL) FDD at 2 GHz data rates, as shown later.
Spectrum MNOS and MNOT in orthogonal bands
• As ISDA is reduced, UAVs are no longer forced to
DL: single-user round robin, whole band
Scheduler connect to far-off dedicated BSs, and can afford reducing
UL: multi-user round robin, 360 kHz each
Downlink power 34 dBW/MHz per beam
their transmission power and interference generated. This
Uplink power Always max power Pmax = 23 dBm results in an increasing SINR for UAVs and GUEs alike.
Beam antenna Circular aperture, HPBW: 4.41◦ , max 30 dBi • Upgrading from ZF to EDA precoding allows both op-
G/T 1.1 dB/K erators to neutralize the increased intercell interference
Users arising from spectrum sharing. For MNOT , this counter-
GUE distribution 15 GUEs per MNOT cell: 80% in buildings measure is key to preserve the legacy GUEs performance.
of 4–8 floors, 20% outdoor at 1.5 m
While not shown for space constraints, similar observations
UAV distribution 1 UAVs per MNOT cell, at 150 m
can be made for the downlink, with two caveats:
eVTOL distribution 0.1–1 eVTOLs per MNOT cell, at 1500 m
Traffic and load Full buffer, fully loaded network • UAVs turn from originators to main victims of inter-MNO
User association Based on RSRP (large-scale fading) interference. Reducing ISDA allows MNOA a correspond-
User antenna Omnidirectional, gain: 0 dBi ing power reduction, which can be used to trade off UAVs
Noise figure 9 dB and GUEs performance.
• The benefits of EDA nullsteering are mostly confined
the strongest cell of their respective serving operators. Both to—and needed by—MNOA . Indeed, the dominant chan-
UAVs and GUEs employ the open-loop power control policy nel eigendirections for both operators correspond to users
specified in 3GPP TR 36.213. The models reported in 3GPP most vulnerable to downlink interference. Intuitively, in
TR 38.901 and TR 36.777 are invoked to characterize the the presence of UAVs, their strong LoS channels domi-
propagation features of all links. nate said subspace and most nulls target receiving UAVs.
We assume the BSs of MNOT and MNOA to be respectively Under the right deployment and interference mitigation
downtilted by −12◦ and uptilted by 45◦ , the former being choices, the dual-MNO paradigm can offer comparable SINRs
commonplace for ISDT = 500 m, and the latter yielding the to a setup where GUEs and UAVs are all served by MNOT .
best UAV performance in most cases. Each cell is equipped However, the spatial and spectrum reuse gains provided by
with an 8 × 8 massive MIMO array of cross-polarized semi- MNOA reflect in the UAV data rates, reported in Fig. 3 for the
directive elements, each connected to a separate RF chain, uplink. These largely benefit from increasing the deployment
resulting in a total of 128 RF chains. For both operators, we density of MNOA and employing EDA precoding. Focusing on
assume perfect channel state information, and consider two the 95%-tile, standalone MNOT with ZF provides 36 Mbps as
different multi-user precoding paradigms: opposed to the 134 Mbps achievable with MNOT -plus-MNOA
• Zero-forcing (ZF) precoding, where each BS spatially and ISDA = 500 m. The former may be sufficient for remote
multiplexes a subset of its users. On one hand, this UAV controlling through HD video, whereas the latter may
paradigm requires low-to-no coordination for radio re- also empower 8K real-time video live broadcast (for future
source allocation since all scheduling, beamforming, and VR applications) and 4×4K AI surveillance (for control and
 the benefits of such an arrangement when offering service to
passengers onboard eVTOLs, flying at 1500 m over an urban
area [5]. Let us define:
• The same operator MNOT as in Example I.
• A satellite operator MNOS , availing of a LEO constella-
tion and operating in an orthogonal S-band (sub-6 GHz).
Each LEO BS of MNOS generates multiple Earth-moving
beams pointing to the ground in a hexagonal fashion, each
creating one corresponding NTN cell [15]. Due to its orbital
movement, the LEO satellite may be seen by the users under
a variable elevation angle, defined as the angle between the
line pointing towards the satellite and the local horizontal
plane, whereby angles closer to 90◦ yield shorter LEO-to-
user distances, and are more likely to be in LoS. Besides the
elevation angle, the NTN performance is affected by the beam
frequency reuse factor (FRF). With FRF = 1, all frequency
resources are fully reused across all beams, whereas with
FRF = 3, they are partitioned into three sets, each reused
every three beams. The assumptions reported in 3GPP TR
38.811 and 38.821 are used to characterize the main NTN
Fig. 2: Uplink SINR for UAVs (top) and GUEs (bottom) with MNOT propagation features.
and MNOA sharing the same spectrum, for ISDT = 500 m and a
variable ISDA , and employing ZF (blue) or EDA (orange) precoding.
This time we focus on the downlink, likely the predominant
ISDA = ∞ denotes all GUEs and UAVs served by standalone MNOT . direction for eVTOL occupants. For the latter, Fig. 4 shows
Solid and transparent bars denote 95%-tile and median, respectively. the CDF of the downlink SINR experienced when all are
served by MNOT and when their traffic is offloaded to MNOS .
For MNOS , various LEO elevation angles are considered. The
following remarks can be made:
• A standalone MNOT employing ZF struggles to guar-
antee coverage to eVTOLs as they proliferate. Indeed,
increasing their number from 0.1 to 1 per cell incurs a
progressively larger outage, i.e., SINR < −5 dB, reaching
up to 18% of the cases (solid black). This is due to
the insufficient angular separation between users, caused
by their density and sheer height, which also renders
nullsteering (not shown) unhelpful.
• Offloading traffic from MNOT to MNOS yields universal
coverage with SINRs ranging between −3 dB and 17 dB
for the elevation angles and beam FRFs considered.
• Moving from FRF = 3 to FRF = 1 entails full reuse—
and thus inter-beam interference—, degrading the median
downlink SINR by approximately 8 dB and 14 dB for
Fig. 3: Uplink UAV rates with MNOT and MNOA sharing the same elevation angles of 90◦ and 87◦ , respectively.
spectrum, for ISDT = 500 m and a variable ISDA , and employing • The SINR experiences a prominent degradation when the
ZF (blue) or EDA (orange) precoding. ISDA = ∞ denotes all GUEs
LEO satellite moves from 90◦ to 87◦ , owing to a larger
and UAVs served by standalone MNOT . Solid and transparent bars
denote 95%-tile and median, respectively. propagation distance and a lower antenna gain, with the
median loss in excess of 8 dB for FRF = 1. Nonetheless,
all offloaded users still remain in coverage, even in the
anti-collision in building-intensive areas, lacking positioning presence of inter-beam interference (FRF = 1).
accuracy) [1]. As for the achievable rates, assuming one eVTOL passenger
per cell over an area of 10.8 km2 —the size of Sant Martí,
B. Example II: Complementing TNs with NTN Infrastructure Barcelona’s business district—yields a total of 150 users,
While primarily targeting underserved areas, NTNs may out of which those in outage (18%, i.e., 27 users) could
also be leveraged to augment urban connectivity, e.g., with be offloaded to MNOS . Under an ideal elevation angle of
MNOT opportunistically leasing spectrum and infrastructure 90◦ and FRF = 3, they would experience median rates of
from a satellite service provider. In this example, we study 3 Mbps. Reducing the density of eVTOLs rapidly increases
 factoring in the interplay of different layers, through dynamic
TN-NTN quality of experience management and scheduling.
Node and device capabilities: By design, GEO satellites
differ from LEO satellites in terms of redundancy mecha-
nisms, antenna designs, transceivers, operational frequency,
and/or internal resources (e.g., storage, processing, and power
availability). The variance in capabilities is yet more apparent
with aerial vehicles, as they are conceived for largely different
purposes and environments, and terminals, whose antennas
range from small and isotropic to active ones capable of
tracking. The above further exacerbates the need for network
management, to guarantee a near-optimal use of radio re-
sources while leveraging this heterogeneity.
Ownership and operations: Mega-constellations are emerg-
ing to expand Internet coverage through hundreds or thousands
of satellites, bringing about frequency coordination and col-
lision avoidance issues, among others. While current systems
lack interoperability, with each operator featuring a vertically
Fig. 4: Downlink SINR for eVTOL passengers when connected to
MNOT and when offloaded to MNOS . For the latter, various LEO
integrated stack, 3GPP standardization will be crucial for
satellite elevation angles and beam FRFs are considered. interconnection, giving way to more heterogeneous scenarios.
With multiple systems designed and operated in an ad-hoc
fashion, their decentralized management and optimization may
their experienced rates as both their absolute number shrinks be a cornerstone to realizing a practical integrated TN-NTN.
and so does the outage percentage from MNOT . Specifically,
0.5 and 0.2 eVTOLs per cell respectively yield 75 and 30 B. Research Directions
eVTOLs in total. Out of these, 8.8% and 2.6% experience
Its extreme heterogeneity makes realizing a 3D network a
SINRs below −5 dB, for a total of 7 and 1 eVTOLs incurring
remarkable endeavor. In the sequel, we propose much-needed
outage, respectively. When offloaded to MNOS , their median
research towards an integrated TN-NTN [1], [7].
rates would be of around 11 Mbps and 80 Mbps, respectively.
While our findings are encouraging, they also suggests that 3D radio access: Next-generation networks will have to
in a future with hordes of high-altitude vehicles, broadband connect flying end-devices at all heights, including their occu-
aerial communications may require higher NTN spatial reuse pants. Our preliminary results vouch for exploiting dedicated
through narrow beams and possibly operating in the Ka-band uptilted cells and NTN platforms to support aerial services.
[4], [15]. This option may be viable for relayed access through Nonetheless, operators will have to seek optimal performance-
a more directive receiver mounted onboard the eVTOL [13]. cost tradeoffs, ensuring coexistence between aerial and legacy
ground users, and between different co-channel technologies.
IV. C HALLENGES AND R ESEARCH D IRECTIONS This goal calls for sophisticated interference management
The availability of TN plus NTN segments is a prerequisite schemes leveraging time, frequency, power, and spatial degrees
for realizing a 3D wireless network. Jointly and optimally de- of freedom, and designed atop realistic air-to-ground channel
signing and operating all platforms and nodes requires further models.
disruptive and interdisciplinary research. In this section, we 3D mobility management and multi-connectivity: Integrated
identify the key obstacles that stand in the way along with the TN-NTN will face the upcoming and unprecedented mobility
most needed technological enablers. challenges brought about both by flying end-devices and
by a mobile infrastructure, dynamically dealing with user
A. The Challenge of Extreme Heterogeneity cell selection, re-selection, and configuration. Beyond current
One chief challenge in realizing an integrated TN-NTN power-triggered procedures, novel use-case-specific and asym-
arises from its extreme heterogeneity, reflected at different metric approaches will be required, also accounting for the
levels as outlined below. handover direction, e.g., within a vertical layer (within a LEO
Radio propagation features: NTNs comprise systems and constellation or inter-HAPS) or across technologies (ground-
end-devices at different altitude layers, each with own service to-air/space or vice versa). Optimal mobility management
features. For instance, GEO satellites provide stable and con- policies will need to trade off reliability, spectral- and energy-
tinuous links to ground devices with a considerable propaga- efficient load balancing, and signaling overhead caused by
tion delay, whereas LEO satellites are characterized by lower- conditional handover preparations and radio link failures.
delay interfaces, but may suffer from service discontinuity 3D network management and orchestration: Meeting the
depending on the constellation density. The type of service heterogeneous and ever more stringent traffic needs across a
provided by each layer must be mapped to the user demand, 3D wireless network will require optimal load distribution,
defining the slices of radio resources to be assigned to each ser- [7] O. Kodheli et al., “Satellite communications in the new space era: A
vice class, accounting for the features of the available TN/NTN survey and future challenges,” IEEE Commun. Surveys Tuts., vol. 23,
no. 1, pp. 70–109, 2021.
radio links, and following their rapidly varying topology. A [8] G. Karabulut Kurt et al., “A vision and framework for the high
service orchestrator should dynamically allocate resources at altitude platform station (HAPS) networks of the future,” IEEE Commun.
NTN nodes according to their availability, mobility patterns, Surveys Tuts., vol. 23, no. 2, pp. 729–779, 2021.
architecture hierarchy, and incoming traffic, ensuring seamless [9] Y. Zeng et al., UAV Communications for 5G and Beyond. Wiley –
IEEE Press, 2020.
service continuity to the end-user in spite of intermittent
[10] W. Saad et al., Wireless Communications and Networking for Unmanned
service link availability and feeder link disruptions. Besides Aerial Vehicles. Cambridge University Press, 2020.
communications, computation and caching resources scattered [11] X. Lin et al., “5G from space: An overview of 3GPP non-terrestrial
across TN and NTN nodes will also need to be optimally networks,” IEEE Commun. Standards Mag., vol. 5, no. 4, pp. 147–153,
2021.
allocated and leveraged.
[12] X. Lin, “An overview of 5G Advanced evolution in 3GPP release 18,”
V. C ONCLUSION arXiv:2201.01358, 2022.
[13] “European Aviation Network,” https://www.europeanaviationnetwork.
In this paper, we connected the dots between ground, aerial, com/en/index.html (accessed July 2022).
and spaceborne communications, and reviewed the key oppor- [14] A. Garcia-Rodriguez et al., “The essential guide to realizing 5G-
tunities and challenges brought about by integrating terrestrial connected UAVs with massive MIMO,” IEEE Commun. Mag., vol. 57,
no. 12, pp. 84–90, 2019.
and non-terrestrial networks. We studied augmenting a ground
[15] J. Sedin et al., “Throughput and capacity evaluation of 5G New Radio
deployment with uptilted cells, and also complementing it non-terrestrial networks with LEO satellites,” in Proc. IEEE Globecom,
with a LEO constellation. We found both to be promising 2020.
avenues for supporting aerial communications, under the right B IOGRAPHIES
design choices: the former entails advanced interference mit-
igation capabilities, the latter hinges on a sufficiently dense
Giovanni Geraci is an Assistant Professor at Univ. Pompeu Fabra in
constellation—to guarantee near-zenith coverage—and a care- Barcelona. He is an IEEE ComSoc Distinguished Lecturer, co-edited
fully designed beam reuse. the book “UAV Communications for 5G and Beyond”, and received
the IEEE ComSoc EMEA Outstanding Young Researcher Award.
R EFERENCES
[1] G. Geraci et al., “What will the future of UAV cellular communications David López-Pérez is an Expert and Technical Leader at Huawei
be? A flight from 5G to 6G,” IEEE Commun. Surveys Tuts., pp. 1–1, Research in Paris. He was a Bell Labs Distinguished Member of
2022. Technical Staff and has co-authored 150+ research articles, 50+ filed
[2] Q. Wu et al., “A comprehensive overview on 5G-and-beyond networks patents, and two books on small cells and ultra-dense networks.
with UAVs: From communications to sensing and intelligence,” IEEE
J. Sel. Areas Commun., vol. 39, no. 10, pp. 2912–2945, 2021. Mohamed Benzaghta is a Ph.D. candidate at Univ. Pompeu Fabra in
[3] F. Rinaldi et al., “Non-terrestrial networks in 5G & beyond: A survey,” Barcelona. He received B.Sc. and M.Sc. degrees from Atilim Univ. in
IEEE Access, vol. 8, pp. 165 178–165 200, 2020. Ankara and his research interests include the integration of terrestrial
[4] M. Giordani and M. Zorzi, “Non-terrestrial networks in the 6G era: and non-terrestrial wireless communications.
Challenges and opportunities,” IEEE Network, vol. 35, no. 2, pp. 244–
251, 2021.
[5] M. Mozaffari et al., “Toward 6G with connected sky: UAVs and beyond,”
Symeon Chatzinotas is Full Professor and Head of the SIGCOM
IEEE Commun. Mag., vol. 59, no. 12, pp. 74–80, 2021. Research Group at SnT, University of Luxembourg, where he is acting
[6] M. M. U. Chowdhury et al., “Ensuring reliable connectivity to cellular- as a PI for more than 20 projects. He was the co-recipient of the
connected UAVs with uptilted antennas and interference coordination,” 2014 IEEE Distinguished Contributions to Satellite Communications
ITU J. Future and Evolving Technol., 2021. Award and has co-authored more than 450 technical papers.