PaperlayerView exact revision on arXiv

Source-linked AI summary

A Survey of Air-to-Ground Propagation Channel Modeling for Unmanned Aerial Vehicles

Wahab Khawaja, Ismail Guvenc, David Matolak, Uwe-Carsten Fiebig, Nicolas Schneckenberger

arXiv:1801.01656v1eess.SP

TL;DR

The paper addresses the limited applicability and coverage of existing AG propagation models for small UAV communications. It surveys measurement campaigns and fading models, identifies their limitations, and outlines research challenges. The survey highlights that UAV-specific modeling must account for distinctive airframe, mobility, environmental, and measurement constraints.

  • Problem

    Small-UAV AG channels are less extensively studied, while existing higher-altitude aeronautical models often do not directly fit low-altitude UAV links.

  • Method

    The paper comprehensively surveys UAV AG channel measurements, propagation characteristics, fading models, limitations, and future research directions.

  • Results

    The survey identifies distinctive UAV AG effects including airframe shadowing, dynamic channel conditions, ground reflections, and practical measurement constraints.

  • Takeaways & Limitations

    Accurate UAV communication-link design requires channel models that reflect UAV-specific mobility, hardware, environmental, and operating conditions.

Abstract

from arXiv · show

In recent years, there has been a dramatic increase in the use of unmanned aerial vehicles (UAVs), particularly for small UAVs, due to their affordable prices, ease of availability, and ease of operability. Existing and future applications of UAVs include remote surveillance and monitoring, relief operations, package delivery, and communication backhaul infrastructure. Additionally, UAVs are envisioned as an important component of 5G wireless technology and beyond. The unique application scenarios for UAVs necessitate accurate air-to-ground (AG) propagation channel models for designing and evaluating UAV communication links for control/non-payload as well as payload data transmissions. These AG propagation models have not been investigated in detail when compared to terrestrial propagation models. In this paper, a comprehensive survey is provided on available AG channel measurement campaigns, large and small scale fading channel models, their limitations, and future research directions for UAV communication scenarios.

I. INTRODUCTION

Small UAVs are expanding across civilian and communication applications, creating demand for accurate air-to-ground channel models tailored to their distinctive operating conditions. This survey reviews measurements, models, limitations, and future research directions.

  • Commercial UAV use is growing across surveillance, filming, disaster relief, goods transport, communication relaying, and recreation.
  • UAVs are expected to support cellular communications as broadband nodes, mobile access points, and components of future 5G systems.
  • Higher-altitude aeronautical channel models generally cannot be applied directly to low-altitude UAV communications.
  • Small UAVs have distinctive airframe shadowing and sharper attitude changes, while their AG channels remain less studied than terrestrial channels.
  • The literature covers payload communications and control/non-payload communications, with payload links commonly using potentially congested unlicensed bands.
  • The survey organizes UAV AG research around channel characteristics, measurements, propagation models, and future challenges.

A. Comparison of UAV AG Propagation with Terrestrial

UAV air-to-ground channels often benefit from stronger line-of-sight propagation than terrestrial channels but can change rapidly with motion and environment. Their multipath, Doppler, and scattering behavior therefore requires UAV-specific characterization.

  • Higher likelihood of line-of-sight propagation can reduce transmit power requirements and improve link reliability relative to terrestrial communications.
  • UAV motion can produce rapidly changing channel statistics and limited spatial stationarity, although hovering or isolated UAVs may experience slower changes.
  • Airframe shadowing, onboard noise, antenna properties, Doppler, and UAV height are important factors in AG channel behavior.
  • AG multipath arises from the ground, terrestrial objects, and sometimes the UAV airframe, with ground reflection often forming the strongest non-LOS component.
  • Geometrically based stochastic models represent scattering objects statistically and can produce intermittent multipath components for moving aircraft.
  • Second-order fading analyses include level-crossing rate, average fade duration, and temporal or frequency correlation functions.

D. Antenna Configurations for UAV AG Propagation

Antenna design and measurement configuration strongly influence UAV AG links because UAVs impose tight spatial, aerodynamic, and computational constraints. Existing campaigns predominantly use simple stationary-endpoint and omni-directional configurations.

  • Antenna number, type, orientation, UAV shape, and material properties all affect AG link performance under UAV space and aerodynamic constraints.
  • SISO and omni-directional antennas dominate measurements, while directional antennas can suffer motion-induced misalignment losses.
  • Multiple antennas can provide diversity or beam steering, but UAV space limits hinder spatial diversity and array processing can require substantial onboard computation.
  • Measurements with both transmitter and receiver moving are rare, despite campaigns spanning L and C bands, varied terrain, and water environments.
  • Antenna mounting commonly places antennas beneath the fuselage or wings to reduce airframe shadowing and aerodynamic disruption.
  • Higher UAV velocity shortens channel coherence time and increases Doppler spread, affecting channel statistics and handover behavior.

B. Challenges in AG Channel Measurements

UAV AG channel measurements face practical constraints from payload, regulations, flight conditions, positioning, power, and limited endurance. These constraints complicate accurate and repeatable characterization.

  • Payload limits, regulated operating height, test cost, restricted transmit power, and onboard processing power constrain AG measurement campaigns.
  • Terrain variation, weather, antenna placement, telemetry requirements, flight time, wind gusts, and GPS accuracy complicate measurement repeatability and distance estimation.

C. AG Propagation Scenarios

UAV air-to-ground measurement scenarios are classified by terrain and cover, with channel sounding adapted to UAV payload and mobility constraints. Hilly or mountainous environments generally have fewer reflections and smaller delay spreads than urban or suburban settings, except when distant strong reflections occur.

  • Measurement approach: UAV channel sounding avoids vector network analyzers because payload, synchronization, and mobility constraints require impulse, correlative, or chirp techniques.The receiver is typically placed on the ground because of payload and processing constraints.
  • Environment classification: Measurement environments are classified as open flat, hilly or mountainous, and over water, with terrain cover such as grass, forest, or buildings affecting channel statistics.The survey uses these three environment classes to organize cited measurement scenarios.
  • UAV types: Fixed-wing UAVs generally fly farther and faster, whereas rotorcraft provide greater agility and can hover or move vertically.These airframe differences shape the measurement scenarios considered for UAV air-to-ground propagation.
  • Open terrain: Urban and suburban areas contain denser, more complex structures than rural areas, and these scatterers strongly influence channel characteristics.Building size, height, density, and composition distinguish the open-terrain subclasses.
  • Hilly/Mountainous: Hilly and mountainous channels generally exhibit fewer reflections and smaller RMS-DS values than urban or suburban channels, unless a distant mountain slope creates a strong reflection.Propagation loss in these areas commonly follows a two-ray model adjusted for surface roughness, with diffraction used around obstructions.
  • Forest environments: Forest measurements with UAVs remain limited, although existing studies analyze tree-volume attenuation and temporal fading for longer-range air-to-ground links.Within forests, channel characteristics are dominated by tree type and density.

4) Water/Sea:

Water and sea channels resemble open-terrain channels but differ through surface roughness and reflectivity, while ground multipath can produce substantial periodic fading. The survey reviews sounding methods and channel statistics while noting that throughput depends strongly on implementation and air-interface parameters.

  • Water/Sea: Over-water path loss can use a two-ray model, with variations caused by surface roughness and occasional large reflections from offshore or shoreline objects.These environments generally have smaller RMS-DS than urban or suburban settings, unless geometry produces significant reflections.
  • Water/Sea: Rough seas can introduce scattering and diffraction for very low-height stations, while frequency- and weather-dependent ducting can reduce loss below free space.Ducting is typically treated statistically because it depends on refractive-index variation with height, frequency, and meteorological conditions.
  • Sounding methods: UAV channel sounding commonly uses short pulses, direct-sequence spread-spectrum, chirps, and multitone or OFDM signals.Short pulses reveal multipath components directly in time but require sufficient pulse energy and can have high peak-to-average power ratios.
  • Measurement parameters: Measurements span 100 MHz to 18 GHz, with many campaigns concentrated near 5 GHz and bandwidths ranging from narrowband to several tens of megahertz or more.One cited campaign used 2.2 GHz ultra-wideband sounding.
  • Channel statistics: Channel impulse responses or time-varying channel transfer functions characterize linearly time-varying air-to-ground channels.These statistics support characterization of propagation channels that may sometimes be approximated as time-invariant.
  • Channel metrics: Throughput is a limited measure of the air-to-ground channel itself because it depends on transmitter and receiver implementations, antenna count, and transmit power.Beam-forming gain, diversity, and capacity are also often estimated for MIMO channels.
  • Path Loss/Shadowing: Path-loss exponents across urban, suburban, hilly, and over-water scenarios are generally close to the free-space value of 2, with fit deviations typically below 3 dB.The cited measurements report scenario-dependent differences while remaining near the free-space exponent.
  • Path Loss/Shadowing: Antenna orientation can change measured path-loss exponents during UAV hovering and movement, complicating removal of antenna-pattern effects from channel estimates.The resulting configuration-specific path-loss model may nevertheless remain useful.

B. Delay Dispersion

The survey characterizes AG channels through delay dispersion, Ricean fading, and Doppler effects. Delay spread depends on environment, while K-factor varies with elevation, frequency, and terrain.

  • Delay-domain measures: PDP measurements estimate delay-domain dispersion, most commonly through RMS-DS, although delay windows and intervals are also reported.The PDP is the power-domain representation of the channel impulse response and may be computed instantaneously or over a WSS spatial volume.
  • Delay-domain measures: Clustered multipath components can be modeled with the Saleh-Valenzuela model, while measured RMS-DS depends on terrain cover.The survey reports PDP measurements across different environments and corresponding RMS-DS statistics.
  • Delay-domain measures: RMS-DS over very low-height sea paths was on the order of a few tens of nanoseconds, with a VHF-to-3 GHz model depending on wave height.The cited sea-propagation model uses finite-difference time-domain analysis of the electric field.
  • Narrowband fading: Ricean K-factor measures dominant-component power relative to all other received components and generally increases with elevation angle.AG channels usually exhibit Ricean amplitude fading because of a line-of-sight component.
  • Narrowband fading: Mean K-factor values were higher in C-band than L-band across urban, hilly or mountainous, and over-water environments.Reported urban means were 12 dB for L-band and 27.4 dB for C-band; over-sea means were 12.5 dB and 31.3 dB, respectively.
  • Doppler effects: Doppler shifts depend on UAV velocity, wavelength, and multipath arrival angle; differing shifts produce Doppler spread and can affect OFDM systems.If different frequency offsets are not mitigated by carrier-frequency synchronization, inter-carrier interference can result.

E. Measured Air Interface Statistics

The survey reviews measured air-interface statistics and channel-modeling approaches for UAV AG links. It distinguishes technology-specific performance measurements from models intended to characterize the propagation channel.

  • Measured performance statistics: Throughput and BER are measured for particular communication technologies and environments but may provide limited information directly relevant to AG channel modeling.The survey identifies both as performance indicators alongside propagation measurements.
  • Simulation studies: Simulation-based channel characterizations imitate real-time environmental scenarios across urban, suburban, sea, hilly, and mountainous settings.Reported simulations span carrier frequencies from 200 MHz to 5.8 GHz and include comparisons of communication systems.
  • Simulation studies: Adjusting UAV altitude can minimize outage probability and maximize coverage area for a given SNR threshold, but increased altitude can also increase path loss.The cited optimization uses the Ricean K-factor's exponential dependence on elevation angle.
  • Model taxonomy: The survey categorizes UAV AG channel models as deterministic, statistical, or hybrid, covering narrow-band, wide-band, and UWB communications.Complete channel models include both large-scale and small-scale effects.
  • Model taxonomy: Deterministic ray-tracing models can be highly accurate but require extensive information about obstacle geometry and electrical properties.Required environmental data include obstacle sizes, shapes, locations, permittivity, and conductivity.
  • Model taxonomy: Quasi-deterministic models represent LOS and earth-surface reflections geometrically while modeling remaining multipath components stochastically from measurement-based distributions.The distributions describe multipath amplitude, delay, and duration for each environment.
  • Model taxonomy: Purely stochastic models may be measurement-free or empirical, but incorporating physical information is reported to improve accuracy.The survey notes that increasing computational capability has made more complex models less prohibitive.
  • Geometry-based stochastic models: Regular-shaped GBSCMs use ellipsoids, cylinders, or spheres and often yield closed forms, whereas irregular-shaped GBSCMs place scatterers randomly and are generally more realistic.Scatterer properties are generally specified beforehand in both classes.

B. Path Loss and Large Scale Fading Models

The survey reviews path-loss and large-scale fading models for mostly-LOS UAV AG channels, including CI, FI, and dual-slope formulations. Model selection depends strongly on propagation geometry, terrain, and available physical references.

  • Large-scale effects: In mostly-LOS AG channels, large-scale fading primarily occurs when an object large relative to a wavelength obstructs the LOS path; models cover path loss and shadowing.When LOS is unobstructed, the remaining large-scale variation is associated with the two-ray earth-surface component.
  • Single-slope models: The close-in model uses free-space path loss at a reference distance, a path-loss exponent, and a shadowing or fit-variation term.The survey identifies the CI model as the most common terrestrial-based formulation in the literature.
  • Single-slope models: Measured path-loss exponents range from approximately 1.5 to 4, while a well-elevated ground-station antenna is expected to yield an exponent near free-space behavior.The free-space path-loss exponent is 2.
  • Single-slope models: The floating-intercept model removes the free-space reference loss and instead uses slope and intercept parameters with a random variation term.Its parameters are represented by α and β, with XFI accounting for path-loss variation.
  • Multi-slope models: Dual-slope path-loss models use different slopes across link-distance ranges when NLOS paths and complex geometries cause large regression errors for single-slope models.The distance ranges are separated by a threshold d1.
  • Model variants: Path-loss models include log-distance, two-ray, shadowing-aware, altitude-dependent, and LOS/NLOS probability-weighted formulations.The cited literature applies these models across diverse environments and UAV heights.
  • Model selection: Model selection is pivotal because the FI slope lacks a standard physical reference and therefore depends on the environment rather than serving as a path-loss exponent.The survey recommends model choices according to terrain cover, openness, geometry, and propagation over water.

C. Airframe Shadowing

Airframe shadowing is a distinctive AG impairment in which aircraft-body obstruction and polarization mismatch attenuate the LOS signal during maneuvers. Its severity depends on roll angle and maneuver duration.

  • Multiple spatially separated antennas can largely, though not always completely, alleviate simultaneous airframe shadowing.
  • Wing-shadowing attenuation generally increases with aircraft roll angle at 970 and 5060 MHz, exceeding 35 dB at both frequencies.
  • Shadowing duration depends on flight maneuver and can exceed tens of seconds during long, slow banking turns.
  • During a banking turn, airframe shadowing and polarization mismatch together produced attenuation exceeding approximately 30 dB on a C-band signal.The example used two bottom-mounted aircraft antennas separated by approximately 1.2 m.

D. Small Scale Fading Models

Small-scale AG fading models represent narrow-band channels or individual multipath components and commonly use Ricean fading, with Rayleigh often fitting NLOS conditions. UAV mobility and intermittent multipath components create rapid, spatially varying channel behavior.

  • Small-scale fading models apply to narrow-band channels, individual MPCs, or tapped-delay-line taps, with fade depth generally varying inversely with signal bandwidth.
  • Time-variant GBSCMs represent small-scale fading with Ricean distributions, while other models describe GA fading using Ricean and Loo distributions.
  • Ricean fading is the most common AG distribution, while Rayleigh typically provides a better fit for NLOS propagation; Nakagami-m and Weibull models may also be used.
  • Intermittent MPCs: Intermittent MPCs arise and disappear along trajectories, changing the channel impulse response and producing wide variation in RMS-DS.
  • Measured PDPs and RMS-DS results show intermittent MPCs as spikes and bumps, illustrating potential rapid time variation in AG channels.

F. MIMO AG Propagation Channel Models

MIMO AG models and measurements examine spatial multiplexing, diversity, and capacity in UAV channels. Gains can be substantial in simulations with favorable conditions, but limited scattering, antenna constraints, and rapidly varying CSI restrict practical benefits.

  • Proper antenna separation and orientation can provide higher spatial multiplexing gains in LOS channels, although UAV mobility makes such alignment difficult.
  • Limited spatial diversity in AG channels permits only moderate capacity gains and may require antenna separations infeasible for small UAVs.
  • Rapidly varying AG channels can make accurate CSI difficult, limiting MIMO gains while adding cost, computational complexity, and power consumption.
  • MIMO measurements observed considerable spatial decorrelation at the ground station from non-planar wavefronts and spatial diversity from UAV antennas at higher elevation angles.
  • Simulations report increased throughput and SNR versus SISO, with capacity increases and lower outage probabilities when perfect instantaneous CSI is available.

G. Ray Tracing Simulations

Ray tracing simulations characterize AG path loss and delay spread across environments by modeling reflections, diffractions, and scatterers. Results generally match analytical two-ray behavior in simple settings, while realistic measurements show additional fluctuations, especially over rough sea surfaces.

  • Ray tracing has been used to characterize UAV AG channels in urban, suburban, rural, and over-sea environments, including 28 GHz and 60 GHz bands.
  • Scatterers add path-loss fluctuations and affect RMS-DS, while UAV height also changes delay spread through the number of significant MPCs reaching the UAV.
  • At 60 GHz, RMS-DS is smaller than at 28 GHz because higher attenuation removes more multipath components.
  • Without scatterers and seawater, the over-sea simulation produces a perfect two-ray path-loss model; nearby buildings add reflected and diffracted MPCs.
  • Measured over-sea path loss fluctuates more than simulations because of equipment variation, ambient noise, and scattering from the rough sea surface.
  • Future research: Future models should cover urban, suburban, industrial, rural, indoor, and quasi-confined environments, including take-off, landing, and strong-multipath conditions.

B. UAV AG Propagation Measurements

UAV AG measurement campaigns must address low-altitude, resource-constrained flight conditions, diverse environments, antenna configurations, onboard disturbances, positioning accuracy, and three-dimensional mobility. The survey also identifies modeling choices ranging from path-loss representations to time-varying and hybrid channel models.

  • Measurement challenges: Existing AG measurements and models mainly target higher-altitude aeronautical links, whereas small UAVs operate at lower altitudes with limited computation, power, and flight duration.These constraints make higher-rate, low-latency, and highly reliable communications challenging for current models.
  • Measurement challenges: Future campaigns should cover varied buildings, reflecting surfaces, difficult landing settings, and diverse UAV-to-UAV velocities and flight situations.Urban UAV-to-UAV channels may require three-dimensional rather than 2.5-dimensional modeling.
  • Measurement implementation: Antenna placement and orientation must limit airframe and motor shadowing and noise, while antenna count depends on frequency, UAV size, and environment.Multiple antennas can improve coverage and diversity but increase computation, space, and power requirements.
  • Measurement implementation: Precise measurements must account for motor and electronics noise, air friction, gusts, temperature changes, and external interference, especially in unlicensed bands.These onboard and environmental conditions affect communication-link characterization for both CNPC and other UAV links.
  • Measurement implementation: GPS supplies position and time references, but its accuracy should be evaluated against measurement requirements and the intended application.The positioning requirement is therefore campaign-specific rather than universally fixed.
  • Channel modeling: UAV channel modeling commonly uses link-distance path loss, but altitude-dependent and indoor models may suit specific settings; accurate models are generally time varying.Time-invariant approximations can apply during hovering over static objects, while future models may incorporate multipath clusters, angular information, parameter correlations, and hybrid GBSCM principles.
  • Survey scope: The survey consolidates UAV AG measurement campaigns, channel statistics, implementation factors, fading and MIMO models, simulations, and future research challenges.Its stated aim is to support development of more elaborate and accurate measurements and channel models.
Loading 1801.01656v1…