Aerial Base Stations for Global Connectivity: Is It a Feasible and Reliable Solution?

Even though achieving global connectivity represents one of the main goals of 5G and beyond wireless networks, exurban areas are still suffering frequent outages because of the lack of proper telecom infrastructures, which are often available only in urban areas. Indeed, cellular network design is usually capacity driven, and thus the densities of base stations (BSs) follow mostly population and especially revenue densities. Contextually, we focus on one of the most promising solutions to provide sufficient and reliable coverage in far-flung areas: aerial base stations (ABSs), which consist of unmanned aerial vehicles (UAVs) carrying cellular BS equipment. In this article, we extensively discuss the problem of bridging what is called the urban–rural digital divide (i.e., the connectivity gap between urban and rural areas) from various perspectives. First, we showcase various alternative solutions and compare conventional terrestrial networks with aerial networks from a techno-economic point of view. Then, we highlight the topological aspects of rural environments and explain how they can affect the actual design of cellular networks. In addition, we investigate both the coverage probability and the reliability of the communication links via simulations, proving that the integration of ABSs can be quite promising in a 6G perspective. Finally, we propose two original extensions of our case study as open problems.


I. INTRODUCTION
According to International Telecommunication Union (ITU), last year roughly 2.9 billion people were still either unconnected or under-connected, and only 27% of the overall population of least developing countries (LDCs) could enjoy the benefits of an Internet connection.On top of that, the consequences of these numbers have also gained importance due to the COVID-19 pandemic.Thus, in the context of the target 9c of the UNs' Sustainable Development Goals (SDGs), wireless networks should aim to offer universal and affordable access to broadband connectivity.
In fact, endowing rural and remote areas with information and communications technologies (ICTs) can be the key to promote communities' cultural and economic growth: countless applications, including precision agriculture, weather monitoring, and online services (for example, in the fields of education, commerce, finance, government, entertainment, and healthcare) depend on the presence of reliable communications systems.
However, the high costs of power, backhaul, installation/maintenance, and security, combined with the low population density and average revenue per user (ARPU) characterizing exurban areas makes them unappealing to telecom providers.Additionally, far-flung areas often lack a reliable power source, which is vital for any communications system, and present challenging topographies.Finally, we recall that a large percentage of telecom infrastructures in small islands and low income exurban areas are powered by diesel generators, which have a high environmental impact and hence cannot be considered as a viable large scale solution to mitigate climate change, as required by the 13 th SDG.
Among various potential solutions, aerial base stations (ABSs) based on drones or other types of aerial platforms, for example airships or gliders, are probably the most promising due to their multiple advantages such as low cost and power consumption, long coverage radius, and high mobility.Nonetheless, several technological challenges (especially in terms of autonomy) still need to be overcome in order to make ABSs commercially viable in most rural areas.

A. Contributions of this Paper
This paper's main contributions are: (i) An original overview of the main technological, economic, and topological aspects of rural connectivity is proposed; (ii) A realistic system setup consisting of terrestrial base stations (TBSs) and efficiently deployed ABSs is introduced, and insightful simulation results in terms of coverage probability and signalto-interference-plus-noise ratio (SINR) meta distribution1 are provided.While our previous paper [1] focused on the coverage analysis of this type of network, to our best knowledge none of the existing literature works investigated the communication reliability via SINR meta distribution in a comprehensive environment including urban and exurban areas; (iii) Two original open problems, which will be introduced and discussed by following the same lines of the previous case study.

B. Outline of this Paper
The rest of this work is structured as follows: Sec.II proposes a concise survey on rural connectivity focusing on alternative solutions and techno-economic aspects of both aerial and conventional terrestrial infrastructures, while Sec III discusses the topological aspects that need to be taken into account when designing and evaluating rural cellular networks.Next, we propose a realistic case study in Sec.IV, and we use the obtained simulation results (in terms of coverage probability and SINR meta distribution) to extract fruitful insights for future rural cellular network design.Finally, we propose our follow-up open problems in Sec.V before concluding the paper in Sec.VI.

II. BRIEF SURVEY ON RURAL CONNECTIVITY
Potential solutions for bridging the digital divide should be tailored to the specific area under consideration and its respective users, while meeting all the regulatory, social, and economic constraints.Contextually, an interesting framework that finely quantifies the digital imbalance has been proposed in [2].Then, the network infrastructure should be sized by estimating multiple factors, such as the current population density and its expected evolution, as well as the percentage of simultaneously active users and their needs in terms of data rate.This section overviews the main alternative technologies and platforms (also displayed in Fig. 1) for achieving global connectivity, and compares conventional terrestrial infrastructures and aerial ones from a techno-economic point of view.In particular, in the first subsection we briefly discuss four paradigms, namely satellites, wind-turbine-mounted base stations (WTBSs), television white space (TVWS), and aerial platforms; The second subsection, instead, focuses on the conventional terrestrial infrastructure and the aerial one, providing details about their components and the respective costs.

A. Alternative Solutions
In this subsection, we concisely overview other potential solutions to bridge the digital divide.
However, for a broader overview of terrestrial, aerial, and space networks in far-flung areas the reader can refer to works such as [3].
1) Satellites: Satellites are considered as a disruptive technology for 5G and beyond networks, especially due to their potential to significantly boost coverage in far-flung areas.Companies such as SpaceX, Amazon, OneWeb, LeoSat, and Boeing are now competing in this new global market by deploying large satellite constellations that cover large portions of the Earth.However, as of today commercial low-Earth-orbit (LEO) satellites cannot provide access directly to cellular user since they require an intermediate dish, which is usually installed on the rooftop of the subscriber's house (even though there are recent collaborations between companies such as Apple, Globalstar, and Qualcomm, or SpaceX and T-mobile, that plan to overcome this issue soon, at least for text messages).Moreover, LEO satellites' trajectory cannot be controlled, and such mega-constellations require advanced equipment for antenna tracking in order to support handover procedures [4].Finally, their environmental impact can become considerable as more and more satellites are launched [5].
2) WTBSs: Exploiting large wind turbines (WTs) for providing exurban cellular connectivity can be an effective solution to combine power and communication infrastructures in one single tower, which compared to a conventional cell tower is endowed of a taller and more robust structure.Whenever conveniently applicable (for example, if existing WTs, government subsidies 2 , or high wind energy potential are available), the proposed solution would in turn bring advantages such as: cost-effectiveness (since the communications transceivers are incorporated into the power infrastructure), continuity of service (since the WT is often connected to a reliable power grid), and high performances (since the robustness of the WTs easily allows to install multi-antenna systems, and considerable altitudes lead to wide coverage radii) [6].Finally, some of the existing WTs are already connected to the core network for transmitting data about temperature, wind speed, and/or humidity to their control centers; therefore, the existing link could also support additional mobile users or sensors.Nonetheless, a considerable percentage of exurban areas is not very suitable for efficient wind energy harvesting.
3) TVWS: The use of unused white space is an attractive solution, since it can allow achieving a coverage radius of tens of kilometers even in presence of obstructions such as trees and thin buildings.Moreover, it benefits from both low cost and short deployment time.One of the main challenges is to coordinate the user devices so that they adjust their transmit power based on the channels available at every instant, so that they do not interfere with TV broadcasters.Databases developed by private companies (e.g., Nominet) can help in computing the required power limits.
It is also important noting that the presence of many underused TV towers strongly supports the practical implementation of this solution.While authors in [7] showed via simulations that TVWS can be effectively combined with high-altitude platforms (HAPs) by using 64-and 256quadrature amplitude modulation (QAM) schemes, an insightful techno-economic analysis about the feasibility of TVWS can be found in [8].
4) Aerial Platforms: Both tethered and untethered aerial platforms can be quite effective in covering rural environments.Indeed, high-altitude untethered platforms such as gliders or airships can provide coverage over tens or even hundreds of kilometers squared (and therefore are appropriate for serving remote areas), while the advantage of tethered platforms lies in their extremely long endurance and the possibility of providing high quality communications thanks to the wired backhaul link (thus they are preferred in suburban environments).As far as we are concerned, a good compromise between these two options is Altaeros' SuperTower, which peculiarity is that it does not need any personnel for deployment.The main advantage of the ST-Flex SuperTower is its rapid deployment, while the ST-300 one is able to carry as much as 300 kg of payload.However, there is also a huge interest in untethered drones, which we will deepen from a novel perspective in the next subsection.
Several works have focused on aerial base stations for rural connectivity.For example, our stochastic-geometry-based analysis in [1] allowed to investigate how the coverage probability behaves as the user gets farther from the town center.In addition, [7] presented a novel architecture where HAPs deliver broadband services via TVWS spectrum.Finally, authors in [9] focused on all the aspects related to untethered drone's endurance when they are operating as ABSs in rural areas.In this paper, instead, we shed light on the techno-economic aspects of ABSs, their integration with the terrestrial infrastructure, and their potential when it comes to enhancing exurban networks' coverage and reliability.

B. Techno-Economic Aspects
On the technical side, the main advantage of the conventional solution is that TBSs have sufficient autonomy and payload capability to host a large number of antennas, which allows to satisfy high capacity demands even for a large number of users, whereas ABSs (and especially untethered drones) are much more limited.In contrast, aerial nodes may easily fly at high altitudes, which allows to take advantage of privileged line-of-sight (LoS) channel conditions over larger areas.Nonetheless, it should be noted that by increasing the altitude, the aerial interference may become excessive, which requires advanced beamforming techniques (as we will assume later on in our case study) or proper cooperation schemes (e.g., based on trajectory and resource allocation optimizations as in [10]).
1) Conventional Infrastructure's Costs: Excluding the overall cost of backhaul (approximately 15 000 USD/km in case of optical fiber [8]), the capital expenditure (CapEx) of a typical rural terrestrial infrastructure is roughly $ 60 000, where equipment, site build, and installation costs respectively account for the 56%, 35%, and 9%.On top of this, an administration cost of 20-30% should be taken into account.In terms of annual operating costs, they account for approximately $ 20 000, fairly distributed among maintenance, site rental, and electricity consumption costs [11,Sec. 3.2.3].Thus, by tackling the problem of global connectivity with the idea of a conventional cellular infrastructure, the expected overall cost of achieving 4G connectivity would amount to $ 388 billion, of which only a negligible part is associated to high income countries, while more than 60% and 30% are required by emerging and low income economies, respectively [12,Sec.V].
2) Aerial Infrastructure's Costs: Since HAPs and tethered platforms are less popular than (untethered) drones, we hereby focus only on the latter.In any case, given the rapid growth of this market, the costs related to aerial infrastructures are expected to decrease considerably.Despite the scarcity of reliable average costs in the literature, we can roughly estimate the cost of purchasing a drone for communications as $ 7200, where: $ 5000 are for the drone, $ 700 are for its battery, $ 1000 are for its power station, and $ 500 are for its flight controller 3 .Furthermore, deploying an aerial network implies the need of a charging infrastructure (for tethered platforms this could be seen as part of the aerial network itself, whereas for HAPs this is less important due to the longer autonomy), whereby issues related to the charging time and the consequent security problem of drones at charging stations, cost of ownership and installation of the stations should be considered.The main options are (i) conventional charging stations and (ii) laser power beaming stations: (i) Studies such as [9] investigated the problem of optimal charging stations' deployment while taking into account the user distribution and the availability of reliable power grids and renewable energy sources.Moreover, advanced technologies (such as the one developed by Asylon Robotics in 2016 [13]) allow to automatically swap the discharged battery with a charged one, therefore limiting the time spent at the charging station to just one minute.Nonetheless, both a larger number of batteries and a more complex system are needed; (ii) Laser power beaming is an advanced technology using high-power (i.e., several kW) laser sources to charge drones.In particular, the energy emitted by laser beams is converted into electric energy by means of a photovoltaic receiver, specifically designed to be sensitive to the wavelengths of the laser sources.Nonetheless, system cost and complexity, as well as safety and overall energy efficiency are important concerns to take into account [14].
Unfortunately, as of today, a detailed economic analysis on charging infrastructures (either conventional charging stations or laser power beaming stations) in rural areas seems missing in literature.Nonetheless, based on our analysis we can fairly consider aerial infrastructures to be 5-10 times cheaper than the conventional ones.

III. TOPOLOGICAL ASPECTS OF RURAL NETWORK DESIGN
Given the economic constraints and the wide range of opportunities, network planning should be done carefully.In this section, some important topological aspects are discussed in order to facilitate the design of rural and far-flung networks.However, it is evident that effective network planning requires collecting big data (e.g., by means of drones or by referring to similar environments).Also, highlands already hosting TV or WT towers could strongly attract investments, since there would be more opportunities to provide ubiquitous coverage.
1) Exurban Area's Size: Estimating the size of the area to cover is evidently fundamental for determining the best solution to deploy and how to scale it.In case of very small areas, it might not be convenient to cover them, unless by relying on the surrounding TBSs.In case of very large areas, instead, it might be complicated to deploy drones because it would require many charging stations (otherwise, an excessive part of drones' stored energy would be spent in moving from the charging station to the desired location), whereas LEO satellites would definitely be more effective.On the other hand, HAPs or towers such as Facebook Connectivity's Supercell are valuable platforms to cover medium-large areas.For intermediate sizes (i.e., just a few km 2 ), we believe that drones would generally be the most convenient solution.
2) Geographical Location: Identifying the type of environment hosting the prospective exurban network is also quite important, because it allows to clarify some important aspects such as: ease of installation and maintenance, presence of power sources, exposure to harsh weather conditions, government regulations, and inhabitants' reactions.In particular, this last aspect strongly depends on the community's perception about the effects of a new wireless technology on its members' privacy and health.ABSs can be the perfect solution for mountainous and hardto-reach areas, where building any terrestrial infrastructure is too expensive and complicated, and part of the satellites' trajectory may be obstructed by a mountain (although in most of the cases said obstruction is not as important to compromise the effectiveness of space networks).
3) Load Distribution: The entity of the load has strong implications on the capacity of the network.Generally speaking, it is recommended to deploy the BSs consistently with the users' density distribution, that is, more BSs close to the users' clusters and less in sparsely populated zones: therefore, ABSs can strongly benefit from their mobility and relocation flexibility whenever the load distribution evolves over time, but their capacity is quite limited due to weight constrains.
Nonetheless, beamforming techniques as well as favorable channel conditions can definitely help in serving users even from relatively far BSs.Finally, note that, in the same environment, there could be multiple types of users with different demand characteristics: for example, rural inhabitants, farmers, and Internet of things (IoT) devices should be considered as three different types of users.
4) State of the Existing Infrastructure (if any): Rural environments usually have a basic cellular infrastructure.As previously mentioned, TV towers and WTs can be endowed of cellular functionalities and integrated to the network architecture.However, sometimes the infrastructure in rural areas is already sufficient (also in terms of backhaul links, power grids, or energy sources), but just needs to be restored or improved, for example, by converting 3G BSs into 4G ones.Nonetheless, using satellites always requires building their own ground stations, which in turn implies additional costs.

IV. NETWORK PERFORMANCES
In this section, we propose insightful simulation results about the joint access-backhaul coverage probability and SINR meta distribution.The coverage probability relates to the chance that a typical user experiences sufficient QoS at a given instant, where the QoS is hereby considered in terms of either SINR, or simply signal-to-noise ratio (SNR).On the other hand, the SINR (or SNR) meta distribution quantifies the probability that a given area is covered for at least a specific fraction x (referred to as reliability) of the time.
To evaluate these performance metrics, we assume the system setup illustrated in Fig. 2, where we have a comprehensive environment that includes urban, suburban, and rural areas; however, when applying the line-of-sight (LoS) model proposed in [15], we consider the parameters of a rural environment.The actual TBSs' density follows a 2D Gaussian distribution centered at the town center and scaled by a factor λ T .On the other side, ABSs are deployed uniformly starting from a distance r e (which identifies a circular ABS exclusion zone A e centered around the town center).
Note that the results are obtained by assuming that the terrestrial antennas devoted to backhaul transmit with very narrow beams.In this way, we can neglect the backhaul interference experienced by any ABS.Anyways, this is already a more general case compared to the one we assumed in [1], where backhaul links were assumed ideal.The main simulation parameters are summarized in Table I, which suggests that NLoS drones will have negligible influence on the user experience because, apart from their longer average distance to the user, they are characterized by a higher path loss exponent and smaller mean loss coefficient.

A. Coverage Probability
As we can see from Fig. 3, the behavior of the joint coverage probability P c as a function of the ABSs' density λ A strongly depends on the distance between the user and the town center (r u ).
In particular, it becomes more and more constant as r u decreases, since urban users evidently have low chances of being in LoS condition with any unmanned aerial vehicles (UAVs).On the other hand, as we move our focus away from the city center we start noticing the need for a higher concentration of ABSs, due to the scarce availability of TBSs.Surprisingly, deploying just 0.04 ABSs/km 2 would be enough for an exurban user equipment (for example, located at distance r u = 12-18 km from the town center) to achieve better coverage than its urban counterpart: this is because the presence of mostly aerial nodes limits the aggregate interference to only its aerial component (nonetheless, it is evident that the aerial interference would become excessive if λ A is strongly increased above the considered range).
Moreover, we can note that the change of behavior with respect to r u is not linear at all: for example, the QoS experienced close to the edge of the exclusion zone is much more constant when going towards the urban area rather than the rural one (in other words, the red curve looks much more similar to the blue one rather than the yellow one).
Finally, note that in our case the backhaul link never represents the bottleneck of the system, but this would not be true (at least for r u ≈ r e ) if the assumption of sharp backhaul beamforming (which implies negligible interferences) was disregarded.that increasing the ABSs' density strongly benefits rural users.However, they also cause a slight decrease of the QoS in the town center.

B. SINR Meta Distribution
Moreover, by comparing the red curves (that is, λ A = 0.04 ABSs/km 2 ), we can state that the town center is characterized by more fairness, meaning that a higher number of links would be able to achieve at least a minimal reliability of just 40%.This can be explained by noting that urban areas generally provide a shorter distance to the closest TBS; nonetheless, the strong aggregate interference almost always prevents from achieving high reliability.

A. Techno-Economic Comparison
An interesting open problem regards the techno-economic analysis when considering conventional cell towers and optical fiber deployment.In particular, it would be interesting to make a fair comparison between various potential technologies for a given economic budget: by considering the technologies and costs mentioned in Sec.II, the costs related to the deployment and operation The difference with the one in Fig. 2 consists in the presence of some WTs distributed as a Poisson cluster process (PCP), where the cluster represents a wind farm.For each wind farm, at most one WT is optimally selected to be mounted with the BS equipment.
of ABSs and their charging infrastructure should be compared with the costs of conventional network architectures providing an equivalent QoS.

B. Including WTBSs
As cellular networks become and more heterogeneous over time, we believe that rural connectivity planning should also take into account the presence of existing or prospective WTs within the area of interest.Therefore, another open problem would be to extend the study proposed in Sec.IV by adding the tier of WTs to the TBSs' and ABSs' tiers.
Furthermore, the performance of the network should be optimized by taking into account all the aforementioned techno-economic constraints.For example, it might be convenient to have just one WTBS per wind farm (see Fig. 5), in order to maintain an acceptable level of interference while improving coverage and capacity.For each wind farm, the BS equipment should be mounted on the optimal WT based on topological aspects such as the distance to the neighbor nodes and/or the user distribution.
Moreover, given the scarcity of TBSs in rural areas, the presence of strategically deployed WTBSs would make it easier to improve the backhaul capability of the network.In other words, WTBSs could allow network operators to serve more users and deploy more ABSs, if needed.

Fig. 1 .
Fig. 1.Main types of platforms for terrestrial, aerial, and space communications for exurban networks.

Fig. 2 .
Fig.2.Considered system setup: the typical user is located at distance ru from the origin and associates to the base station (BS) that provides the maximum average received power.All the ABSs are placed outside a circular exclusion zone Ae .

Fig. 3 .
Fig. 3. Coverage probability as a function of the ABSs' density for various values of ru .

Fig. 4
Fig.4shows interesting results about the SINR meta distribution at the town center and at 18 km of distance from it.Consistently with Fig.3, we can clearly notice from the solid lines

Fig. 4 .
Fig. 4. SINR meta distribution for typical urban and rural users, for various values of λA .

Fig. 5 .
Fig. 5. System setup for the proposed open problem.The difference with the one in Fig.2consists in the presence of some WTs distributed as a Poisson cluster process (PCP), where the cluster represents a wind farm.For each wind farm, at most one WT is optimally selected to be mounted with the BS equipment.