PaperlayerView exact revision on arXiv

Source-linked AI summary

Cellular Connectivity for UAVs: Network Modeling, Performance Analysis and Design Guidelines

Mohammad Mahdi Azari, Fernando Rosas, Sofie Pollin

arXiv:1804.08121v1cs.NIcs.PF

TL;DR

The paper asks whether cellular networks can reliably serve UAVs alongside ground users and how aerial users affect network efficiency. It develops a generic analytical framework for coverage, throughput, and area spectral efficiency, then derives design guidelines. Results show that antenna configuration and tier selection can substantially improve UAV links, while UAV inclusion benefits moderate-density networks but can hurt ultra-dense-network performance.

  • Problem

    The paper addresses whether cellular networks can provide aerial connectivity, which factors limit UAV performance, and how aerial and ground users can coexist efficiently.

  • Method

    The paper develops a generic LoS/NLoS cellular-network model and derives exact and approximate expressions for UAV coverage probability, achievable throughput, and area spectral efficiency.

  • Results

    Optimal UAV antenna tilting raises coverage from 23% to 89% and throughput from 3.5 b/s/Hz to 5.8 b/s/Hz, while optimal tier selection raises 100 m coverage from 10% to 80% and throughput from 0.25 b/s/Hz to 4 b/s/Hz.

  • Takeaways & Limitations

    UAV integration can improve area spectral efficiency at moderate BS density but can reduce ultra-dense-network gains, making antenna, altitude, tier, and density choices important.

Abstract

from arXiv · show

The growing use of aerial user equipments (UEs) in various applications requires ubiquitous and reliable connectivity for safe control and data exchange between these devices and ground stations. Key questions that need to be addressed when planning the deployment of aerial UEs are whether the cellular network is a suitable candidate for enabling such connectivity, and how the inclusion of aerial UEs might impact the overall network efficiency. This paper provides an in-depth analysis of user and network level performance of a cellular network that serves both unmanned aerial vehicles (UAVs) and ground users in the downlink. Our results show that the favorable propagation conditions that UAVs enjoy due to their height often backfire on them, as the increased co-channel interference received from neighboring ground BSs is not compensated by the improved signal strength. When compared with a ground user in an urban area, our analysis shows that a UAV flying at 100 meters can experience a throughput decrease of a factor 10 and a coverage drop from 76% to 30%. Motivated by these findings, we develop UAV and network based solutions to enable an adequate integration of UAVs into cellular networks. In particular, we show that an optimal tilting of the UAV antenna can increase their coverage and throughput from 23% to 89% and from 3.5 b/s/Hz to 5.8 b/s/Hz, respectively, outperforming ground UEs. Furthermore, our findings reveal that depending on UAV altitude, the aerial user performance can scale with respect to the network density better than that of a ground user. Finally, our results show that network densification and the use of micro cells limit UAV performance. While UAV usage has the potential to increase area spectral efficiency (ASE) of cellular networks with moderate number of cells, they might hamper the development of future ultra dense networks.

I. INTRODUCTION

The paper investigates whether cellular networks can reliably serve UAVs alongside ground users and how aerial users affect network efficiency. It develops an analytical framework and design guidelines addressing propagation, interference, antenna configuration, tier selection, and network density.

  • I. INTRODUCTION: Cellular networks are promising for UAV connectivity, but infrastructure optimized for ground users may impose antenna-gain and interference penalties on aerial users.Downward BS antenna tilt reduces aerial-user gain, while ground-oriented spatial reuse and interference assumptions may not apply to UAVs.
  • I. INTRODUCTION: The study asks whether cellular networks can serve aerial users, which factors limit their performance, whether UAV design can improve connectivity, and what coexistence trade-offs arise.
  • I. INTRODUCTION: The framework models UAV cellular performance using coverage probability, achievable throughput, and area spectral efficiency under LoS/NLoS propagation and UAV antenna beamwidth and tilt.
  • I. INTRODUCTION: Omnidirectional UAVs suffer interference-driven degradation, whereas environment-dependent antenna tilt, beamwidth, and altitude can improve coverage and throughput, with different optima across metrics.
  • I. INTRODUCTION: Optimal tier selection improves UAV link coverage from 10% to 80% and throughput from 0.25 b/s/Hz to 4 b/s/Hz at 100 m altitude.
  • I. INTRODUCTION: UAV inclusion can improve network area spectral efficiency at moderate density but becomes detrimental in ultra-dense networks.

B. Channel Model

The channel model separates LoS and NLoS propagation, fading, and blockage effects in an urban cellular environment. LoS probability increases with UAV altitude and decreases stepwise with horizontal distance.

  • B. Channel Model: The model represents LoS and NLoS links with separate path-loss exponents and reference-distance constants.
  • B. Channel Model: Nakagami-m fading models small-scale channel variation, with unit-mean Gamma fading power and integer fading parameters for analytical tractability.
  • B. Channel Model: LoS probability increases with UAV altitude and decreases as a step function of ground distance.
  • B. Channel Model: Received power combines transmit power, path loss, fading power, and the total transmitter–receiver antenna gain.
  • B. Channel Model: Urban blockages are modeled using building area fraction, building density, and Rayleigh-distributed building heights.
  • B. Channel Model: Independent link LoS probabilities decompose the UAV-visible BS process into inhomogeneous LoS and NLoS Poisson subprocesses.

D. User Association Strategy and Link SINR

Users associate with the BS providing the highest SINR, and link performance is evaluated through coverage probability, throughput, and area spectral efficiency. These metrics capture reliability, raw rate, and network-level data rate, respectively.

  • D. User Association Strategy and Link SINR: The serving BS is selected by highest SINR and may be LoS or NLoS rather than the geographically closest BS.
  • D. User Association Strategy and Link SINR: Aggregate interference consists of contributions from other LoS and NLoS BSs within the UAV antenna’s visible region.
  • D. User Association Strategy and Link SINR: The instantaneous link SINR combines received serving power, aggregate interference, and receiver noise power.
  • D. User Association Strategy and Link SINR: Coverage probability depends on UAV altitude and SINR threshold and measures whether the link satisfies a target requirement.
  • D. User Association Strategy and Link SINR: Throughput depends on altitude and network density and complements coverage probability by measuring average raw throughput.
  • D. User Association Strategy and Link SINR: Area spectral efficiency measures achievable data rate per square meter while accounting for the fraction of aerial users sharing network resources.

3) Area Spectral Efficiency (ASE):

The paper derives exact coverage expressions and supporting approximations, then evaluates throughput and area spectral efficiency for UAV–ground-user coexistence. The analysis is intended to quantify link and network effects of aerial users.

  • 3) Area Spectral Efficiency (ASE):: Area spectral efficiency quantifies network-level achievable data rate per square meter and accounts for aerial and ground users served by the network.
  • 3) Area Spectral Efficiency (ASE):: The analysis derives an exact coverage-probability expression, proposes approximations for numerical evaluation, and calculates user throughput and network area spectral efficiency.
  • 3) Area Spectral Efficiency (ASE):: The coverage theorem characterizes serving-BS distance and aggregate interference through the serving-distance distribution and an interference Laplace transform.
  • 3) Area Spectral Efficiency (ASE):: The exact coverage framework can also evaluate omnidirectional antennas by setting the directional-antenna geometry parameters to their omnidirectional values.

B. Approximations for UAV Coverage Probability

The paper simplifies UAV coverage analysis by neglecting NLoS links and noise, then approximates interference statistics with a Gamma distribution to obtain more tractable coverage expressions.

  • The approximation is motivated by UAVs' higher LoS probability, which makes nearby LoS BSs dominate received power over NLoS BSs.The same conditions support neglecting noise because aggregate interference is high.
  • Proposition 1 simplifies UAV coverage probability by eliminating NLoS links and thermal noise effects.The simplification follows from setting the corresponding NLoS and noise terms approximately to zero.
  • The approximations remove NLoS-related terms from the Laplace-transform calculation, with their accuracy evaluated later.
  • 2) Moment Matching:: Gamma-distributed interference statistics obtained by moment matching yield a closed-form approximation of the interference Laplace transform.This removes the associated integral and derivatives from coverage calculations.
  • For φt = 0, the first and second interference moments have specialized expressions corresponding to a downward-pointing or omnidirectional UAV antenna.

IV. SYSTEM DESIGN: STUDY CASES AND DISCUSSION

The study uses network simulations and numerical evaluations to examine UAV connectivity design parameters, compare aerial and ground users, and assess heterogeneous networks and densification.

  • The study runs network simulations and numerical evaluations using parameter values listed in Table I.
  • IV-A/B: The evaluation validates coverage expressions and approximations, studies system-parameter effects on UAV coverage and throughput, and compares UAVs with ground users.
  • IV-C/D: The study also examines UAV connectivity in heterogeneous networks and under network densification.

A. Analysis of Accuracy

The analytical coverage results agree closely with simulations, while the approximations track UAV coverage trends; increasing altitude concentrates UAV SINR distributions but does not prevent higher-altitude coverage loss with omnidirectional antennas.

  • The UAV approximations follow the observed trend in exact coverage probability, whereas they are not provided for ground users because NLoS links are prominent.
  • At higher altitude, omnidirectional UAV coverage drops significantly.
  • 105 network realizations produce simulation results in perfect correspondence with Theorem 1 and support the accuracy of Propositions 1 and 2.
  • As UAV altitude increases, the SINR distribution becomes more concentrated because total LoS signal power grows and dominates multipath scatterers.
  • For ground users, multipath scatterers remain prominent in determining the SINR distribution because their LoS probability with BSs is low.

B. Design Parameters

Cellular-connected UAV performance depends strongly on altitude, environment, antenna beamwidth and tilt, and network density. Interference creates key trade-offs: optimized antenna designs can improve coverage and throughput, but densification and excessive altitude can degrade UAV links.

  • Impact of Altitude: An optimum UAV altitude maximizes cellular-link performance by improving the serving-BS LoS probability while keeping most interfering BSs in NLoS conditions.At higher altitudes, more interfering BSs become LoS and reduce communication-link quality.
  • Impact of Altitude: 10% of maximum throughput is achieved by an urban UAV at hu = 100 m, compared with maximum throughput at hu = 10 m.The feasible operational altitude range is therefore limited for an omnidirectional UAV.
  • Impact of the Environment: Coverage degradation from ground level to altitude is more severe in less obstructed environments: suburban coverage falls from 90% to 4% at 150 m, versus 76% to 10% in urban areas.Ground users have the best suburban coverage, while UAVs at 50 m or higher are best served in urban scenarios.
  • Impact of UAV Beamwidth: An optimal beamwidth balances increased BS visibility and coverage against the aggregate interference caused by a wider beam.The beamwidth maximizing coverage probability differs from the one maximizing throughput, creating a throughput-coverage trade-off.
  • Network Tier and Density: Optimal tier connectivity is altitude-dependent, with macro cells superior to micro cells at high altitudes and a crossover point determined by the performance metric and antenna configuration.From the UAV perspective, network performance converges to zero as the network becomes dense, although lowering UAV altitude partially mitigates the decline.
  • Impact of UAV Antenna Tilt: At λ = 5 BSs/Km2, optimal antenna tilt raises link coverage from 23% to 89% and channel capacity from 3.5 b/s/Hz to 5.8 b/s/Hz.Tilting benefits sparse to medium dense networks, but becomes ineffective as densification increases interfering BSs and the optimum tilt converges to zero.

C. Heterogeneous Networks - Tier Selection

The paper analyzes UAV connectivity across heterogeneous macro- and micro-cell networks and shows that tier selection, antenna design, altitude, and densification jointly determine performance. Appropriate tier selection and UAV design can improve coverage and throughput, while excessive densification can reduce both user and network gains.

  • Tier selection: Macro and micro cells create a density–interference trade-off for UAV connectivity because macro cells are less dense but have higher transmit power and generate more interfering LoS links.The model evaluates UAV performance using BS density λ, height h_b, and transmit power P_tx.
  • Tier selection: An altitude-aware tier selection strategy can enhance UAV link coverage and throughput by switching between micro and macro cells according to altitude.Lower-altitude UAVs are better served by micro cells, whereas higher-altitude UAVs are better served by macro cells.
  • Tier selection: At 100 m, selecting the appropriate tier increases coverage probability from 10% to 80% and provides 15 times throughput enhancement for (φB, φt) = (150o, 0o).The switching altitude depends on antenna beamwidth, tilt angle, and whether coverage or throughput is optimized.
  • Network densification: UAV throughput is maximized at a finite BS density and then decreases to zero as the network is further densified.The optimum density for maximum performance decreases as UAV altitude increases.
  • Framework: A generic framework derives closed-form user and network performance metrics to provide design insights for integrating UAVs into cellular networks.The framework supports analysis of coverage probability, achievable throughput, and area spectral efficiency.
  • Network densification: Adding UAVs can increase area spectral efficiency at moderate BS density, but in ultra-dense networks their inclusion becomes detrimental to network performance.The scaling analysis identifies signal-dominated, interference-dominated, and balanced regimes for aerial users.

APPENDIX A PROOF OF THEOREM 1

Appendix A derives the conditional coverage formulation by characterizing LoS and NLoS serving regions and integrating over their geometry. The proof uses spatial probability distributions, network geometry, and the PPP probability generating functional.

  • Serving-link distributions: The serving-distance distributions f_L_RS(rS) and f_N_RS(rS) describe the serving BS distance under LoS and NLoS conditions.Regardless of the serving condition, interference comes from both LoS and NLoS BSs.
  • Coverage formulation: The coverage expression combines unconditional LoS-BS density with probabilities that no stronger LoS or NLoS BS exists.The relevant parameters and integrals are specified through Theorem 1.
  • Serving-link regions: The proof defines noN(rS) as locations where an NLoS signal is stronger than the LoS signal at rS.The LoS and NLoS exclusion regions depend on network geometry.
  • Geometric characterization: The proof represents the UAV coverage region C as an elliptical section determined by beamwidth φB and tilt angle φt.Its semi-major axis, semi-minor axis, and origin are used to characterize the region geometrically.
  • Geometric characterization: The LoS serving region is obtained from the intersection of C with a disc of radius rS centered at the origin.The corresponding angular limits ϕ1(r) and ϕ2(r) follow from geometry.
  • Coverage formulation: The interference condition is evaluated through an inequality involving N0 + I > T, with gamma-distributed desired fading and the PPP probability generating functional used in subsequent processing.These steps transform the interference and coverage terms into the theorem’s final expression.

APPENDIX B PROOF OF LEMMA 1

Appendix B derives interference moments from the interference Laplace transform. It then applies the resulting relationships to obtain the desired variance expression.

  • Interference moments: The first moment of interference I is calculated from its Laplace transform.
  • Interference moments: The variance of I is obtained using a relationship applied after the first-moment derivation.The appendix states that a similar derivation yields the desired result.

APPENDIX C PROOF OF COROLLARY 1

Appendix C specializes the angular integration to zero UAV antenna tilt and obtains the corresponding corollary. The derivation uses equal angular limits and an analogous conditional-interference argument.

  • Zero-tilt specialization: Assuming φt = 0 makes the angular limits equal, ϕ1(r) = ϕ2(r) = π/2, simplifying the integral in (21).
  • Zero-tilt specialization: The conditional interference term I|RS is obtained using a similar rationale after the zero-tilt integral is simplified.
Loading 1804.08121v1…