A Survey of Channel Modeling for UAV Communications

Aziz Altaf Khuwaja; Yunfei Chen; Nan Zhao; Mohamed-Slim Alouini; Paul Dobbins

Source-linked AI summary

A Survey of Channel Modeling for UAV Communications

Aziz Altaf Khuwaja, Yunfei Chen, Nan Zhao, Mohamed-Slim Alouini, Paul Dobbins

arXiv:1801.07359v1Jan 23, 2018eess.SP

TL;DR

UAV channels have distinctive air–ground and air–air behavior, including non-stationarity and airframe shadowing, making established terrestrial and satellite models insufficient. This paper surveys low-altitude measurement campaigns and contemporary empirical and analytical modeling approaches, then identifies important unresolved channel-modeling challenges.

Problem
Established terrestrial and satellite channel models are often unsuitable for UAV propagation, whose spatial–temporal non-stationarity and airframe shadowing remain insufficiently characterized.
Method
The paper surveys low-altitude UAV measurement campaigns and reviews empirical and statistical channel models for UAV channel characterization.
Results
The survey categorizes low-altitude campaigns by sounding and communication platform and reviews empirical models for air–ground and air–air propagation.
Takeaways & Limitations
UAV channel characterization requires attention to low-altitude measurements, non-stationarity, and airframe-shadowing effects.

Abstract

from arXiv · show

Unmanned aerial vehicles (UAVs) have gained great interest for rapid deployment in both civil and military applications. UAV communication has its own distinctive channel characteristics compared with widely used cellular and satellite systems. Thus, accurate channel characterization is crucial for the performance optimization and design of efficient UAV communication systems. However, several challenges exist in UAV channel modeling. For example, propagation characteristics of UAV channels are still less explored for spatial and temporal variations in non\textendash stationary channels. Also, airframe shadowing has not yet been investigated for small size rotary UAVs. This paper provides an extensive survey on the measurement campaigns launched for UAV channel modeling using low altitude platforms and discusses various channel characterization efforts. We also review the contemporary perspective of UAV channel modeling approaches and outline some future research challenges in this domain.

I. INTRODUCTION

UAV channels differ from terrestrial and satellite channels through distinct air–ground and air–air links, non-stationarity, and airframe shadowing. This survey reviews measurement campaigns, channel models, and research challenges for low-altitude UAV communications.

Channel characteristics: UAV communication channels have distinct air–ground and air–air links, spatial and temporal non-stationarity, and airframe shadowing.These characteristics distinguish UAV communications from conventional wireless systems.
Motivation: Terrestrial and satellite channel models are often poorly suited to UAV propagation because of these unique channel attributes.Reliable models are needed for wireless-technique evaluation and link-budget calculations.
Modeling approaches: Analytical UAV channel models include deterministic, tapped delay line, and geometric-based stochastic approaches.They address environmental propagation and fading, wideband multipath, and spatial–temporal characteristics, respectively.
Measurement campaigns: Analytical models may fail to describe real propagation behavior because their assumptions are insufficiently realistic.Measurement campaigns are therefore essential for empirical characterization.
Research scope: High-altitude manned-aircraft findings cannot be directly applied to single-hop UAV networks operating at low altitude.The survey emphasizes low-altitude platforms, including operations up to 120 m under cited regulations.
Survey contribution: The survey covers low-altitude measurement campaigns, empirical and analytical channel models, airframe shadowing, non-stationarity, diversity, and future research challenges.It specifically addresses underexplored shadowing, WSSUS limitations, and Doppler characterization.

II. MEASUREMENT CAMPAIGNS

Measurement campaigns are necessary because analytical models may rely on unrealistic assumptions. The survey organizes UAV measurements by platform, wireless technique, and channel-sounding approach, covering large- and small-scale fading and diversity.

Rationale: Field measurements are needed to understand real UAV propagation behavior when analytical assumptions are inadequate.Campaigns examine diverse environments and support empirical channel models.
Characterization outcomes: Reported campaigns investigate AG path loss, fading, Doppler, altitude effects, antenna orientation, and diversity across urban, suburban, open, and wooded environments.The reviewed measurements include both large-scale and small-scale channel behavior.
UAV platforms: Manned-aircraft campaigns are expensive and difficult, motivating UAV-based measurements with onboard processing, wireless equipment, antennas, GPS, and IMUs.The instrumentation records both RF measurements and flight dynamics such as pitch, yaw, and roll.
Campaign categories: The survey groups UAV measurement campaigns into channel sounders, IEEE 802.11 radios, and widely deployed cellular infrastructure.These groups span narrowband and wideband sounding as well as operational communication systems.
Narrowband systems: Narrowband studies characterize frequency-nonselective fading but cannot resolve individual multipath components for rich-multipath, coherence-bandwidth, or MIMO-capacity analysis.Their limited sounding spectrum restricts frequency-selective characterization.

2) Wideband Measurement Systems:

Wideband measurement systems capture frequency-selective UAV channel behavior, including impulse responses, delay spreads, power-delay profiles, and Doppler spectra. They provide richer characterization but require greater computational resources and impose cost and size constraints.

Wideband capability: Wideband sounders derive frequency-selective parameters from channel impulse responses and support multipath characterization.Correlative sounders transmit pseudonoise sequences and correlate the received signal with the same sequence.
Measurement outputs: USRP-based campaigns measure channel impulse responses, delay spreads, power-delay profiles, cumulative distributions, and altitude-dependent multipath behavior.One study observed random multipath-component behavior at different UAV altitudes.
Campaign settings: Wideband campaigns have examined AG propagation using LTE, UWB, and SDR platforms across open, suburban, residential, wooded, and desert settings.The reviewed setups vary frequency, altitude, receiver placement, and terrain.
Observed statistics: Desert terrain produces greater AG delay spread than residential terrain, while RMS Doppler spread follows a log-normal trend.These findings come from CDF-based channel-statistics analyses.
Trade-offs: Wideband systems require additional computation for raw-data processing and may be unsuitable for real-time fading characterization.Equipment cost and physical dimensions are additional constraints.

B. IEEE 802.11 based UAV Measurements

IEEE 802.11 measurements provide low-cost, flexible characterization of UAV channels, including altitude, antenna orientation, path loss, fading, and throughput. Results commonly show free-space-like path loss while revealing orientation- and altitude-dependent effects.

Platform rationale: Commercial IEEE 802.11 radios are useful for studying altitude-dependent multipath and antenna-orientation effects in UAV channels.Their low power consumption, cost effectiveness, and flexibility support small-UAV integration.
Limitations: IEEE 802.11 measurements can be affected by interference, fixed narrowband frequency, and limited communication range.Maintaining high physical-layer SINR is identified as one possible response to adjacent-radio interference.
Orientation effects: 170°–230° produced good signal reception, whereas 240°–260° produced the worst signal in one yaw-angle measurement.The experiment examined ground-reflected multipath and orientation-dependent reception.
Fading models: A height-dependent Rician model was proposed with K factor reliant on UAV altitude.The model represents ground-reflected multipath over flight altitudes between 10 and 40 m.
Channel statistics: For both AG and AA channels, log-distance path-loss exponents roughly matched free-space propagation, while Nakagami m fit multipath fading.These results were obtained from measurements using multiple horizontally aligned antennas.
Antenna configuration: Optimal antenna orientation can reduce altitude effects on received signal strength and throughput, while horizontal alignment reduces yaw-angle effects on throughput.Open-field propagation also followed free-space behavior.

C. Cellular–Connected UAV Measurements

Cellular-connected UAV measurements assess aerial performance across urban, rural, and suburban environments, while revealing altitude-related coverage limitations and propagation anomalies. These campaigns motivate altitude-aware and more robust channel models before widespread deployment.

Measurement campaigns: Cellular networks were evaluated for UAV communication using LTE, GSM, and UMTS measurements across multiple environments and flight platforms.Campaigns used quadcopters, fixed-wing UAVs, captive balloons, and meteorological balloons in locations including San Diego, China, Germany, Lisbon, Denmark, Australia, and California.
Measurement findings: Higher UAV altitude reduced base-station availability in one rural campaign, although rural coverage was better than urban coverage because of fewer ground obstacles.
Measurement findings: A cellular measurement study found received power could drop when the UAV climbed above the base-station height because of the base-station antenna radiation pattern.
Measurement findings: Altitude increases reduced path-loss exponents and shadowing variation, demonstrating that UAV channel models require altitude-dependent parameters.
Measurement findings: A compositional path-loss model identified low-coverage “holes in the sky” beyond line of sight, associated with two-ray reflections, diffraction losses, and antenna-pattern nulls.
Open challenges: Cellular infrastructure offers broad coverage through handover, but down-tilted sector antennas limit air–ground propagation above base-station height and infrastructure failures threaten disaster-response use.

III. EMPIRICAL CHANNEL MODELS FROM MEASUREMENT CAMPAIGNS

Measurement campaigns investigate how UAV motion, geometry, antenna orientation, and environment shape rapidly varying air–air and air–ground channels. Their empirical results show diverse path-loss and fading behavior, while unified conclusions remain unresolved.

Measurement-driven modeling: UAV cruising creates spatial–temporal channel variations, so measurements relate channel parameters to altitude, elevation angle, separation distance, and environment.
Measurement-driven modeling: No unified channel-modeling conclusions have yet been established despite extensive measurement efforts.
Air–air channels: Air–air channels generally show lower path-loss exponents than air–ground channels, including reported air–ground and ground–air exponents of 2.32 and 2.51.
Empirical variation: Reported air–air and air–ground path-loss exponents varied substantially across campaigns, including 0.93 versus 1.50 and 1.92 versus 2.13, respectively.
Empirical variation: Air–air propagation depends on vertical and horizontal separation in line-of-sight conditions, while beyond-line-of-sight operation can incur significant attenuation and raises Doppler concerns at higher velocities.

B. Air–Ground (AG) Channel Characterization

Air–ground measurements show that altitude, distance, and geometry strongly influence path loss and shadow fading. Models therefore incorporate three-dimensional flight parameters, environmental effects, and mechanisms such as reflections, diffraction, and antenna-pattern nulls.

Large-scale fading: Above a breakpoint altitude, path loss becomes roughly similar to free-space propagation as the first Fresnel zone becomes sufficiently cleared.
Model parameters: Air–ground models variously account for distance, frequency, altitude, antenna tilt, elevation angle, foliage, shadow fading, and urban or rural terrain.
Coverage mechanisms: A compositional model identified two-ray ground reflections, diffraction losses, and antenna-radiation nulls as predominant causes of low coverage zones.
Large-scale fading: Air–ground large-scale fading is dominated by UAV altitude and distance from the ground, requiring realistic models to represent these parameters in three-dimensional coordinates.
Open challenges: Further characterization of antenna design and orientation is needed to improve UAV communications.

2) Small–Scale Fading Statistics:

The survey reviews statistical models and measurements used to characterize UAV small-scale fading, including temporal variation, multipath, and dispersive effects. Nakagami and Rician models are commonly effective, while spatial variation remains less studied.

Small-scale fading characterization examines multipath propagation and fading behavior because these effects influence UAV communications.
Rician and log-normal models characterize received-signal variations, with LOS and diffuse multipath components represented statistically.
Nakagami m modeling uses measured multipath components from channel impulse responses, with parameters evaluated empirically across time-delay bins.
m > 1, and Nakagami m fading provided the best fit when theoretical Rayleigh and Nakagami distributions were compared.
0.06µs and 28.96 Hz were the mountainous-desert median RMS delay spread and Doppler frequency spread, compared with 0.03µs and 28.06 Hz residential values.The larger desert delay spread was attributed to rough mountainous scatterers along the flight path.
Small-scale studies mostly address temporal variation, while spatial variation remains comparatively understudied across reported air-to-ground cases.

IV. ANALYTICAL CHANNEL MODELS

Analytical UAV channel models characterize propagation using deterministic, stochastic, and geometry-based stochastic approaches. Existing studies use environmental information, ray tracing, and statistical propagation conditions to derive path-loss, shadowing, coverage, and transmission-rate models.

Analytical models characterize propagation stochastically under specified assumptions and parameters, supporting communication-system performance prediction.
Deterministic models use environment-specific layouts and databases, while stochastic models rely on channel statistics without requiring specific location information.
Geometry-based stochastic models assume random environmental scatterers to obtain spatial-temporal statistics.
3D ray-tracing studies modeled urban air-to-ground propagation from 200 MHz to 5 GHz and aerial altitudes of 100 to 2000 m.
Elevation angle significantly affected excess path loss in low-altitude-platform simulations that grouped LOS and NLOS channel conditions.
Analytical frameworks extended free-space path loss for LOS and NLOS conditions and supported optimization of radio coverage probability and maximum transmission rate.

B. Stochastic Channel Model

Stochastic UAV channel models represent multipath, delay, Doppler, and spatial–temporal behavior through statistical and geometry-based formulations. Their accuracy depends on modeling assumptions and, for non-stationary channels, estimating the stationary interval.

Stochastic models: Stochastic models use tapped-delay lines to represent LOS, NLOS, and intermittent multipath components with statistical delay, amplitude, phase, and Doppler descriptions.Reported models include nine-tap lines and formulations for intermittent-component existence, excess delay, duration, and time-varying Doppler shifts.
Model accuracy: The accuracy of stochastic models depends on estimating the stationary interval in the non-stationary UAV channel.This dependency is especially relevant when model parameters are derived from fading statistics of the channel impulse response.
Stochastic models: Wideband stochastic models characterize UAV air–ground channels using measured data from different environments and estimated stationary intervals.One set of measurements used a stationary interval of 15 m at the C band, while an over-water model combined two-ray propagation with an intermittent third ray.
Geometry-based stochastic models: Geometry-based stochastic models derive narrowband and wideband statistics from three-dimensional scattering geometries for antenna-array and diversity-system design.These models analyze quantities including delay, gain, phase, angle of arrival, space–time correlation, and Doppler characteristics.
Geometry-based stochastic models: Terrain morphology affects the Doppler–delay spread spectrum when analytical results are compared with terrain-based digital-elevation-model simulations.This links channel spectral behavior to the surrounding terrain rather than treating propagation geometry as environment independent.
Geometry-based stochastic models: Geometry-based models often rely on environment-specific scatterer assumptions, including ellipsoidal, cylindrical, uniformly distributed, or scatter-free airborne regions.Such assumptions support analytical expressions for propagation and correlation but make model behavior dependent on the surrounding physical environment.

V. IMPORTANT ISSUES

The survey identifies airframe shadowing and channel non-stationarity as important UAV-specific issues, while reviewing measurement-based findings on propagation and diversity. Existing evidence also shows performance gains from multiple antennas, but experimental coverage remains limited.

A. Airframe Shadowing: Aircraft structures and banking maneuvers can obstruct the direct line-of-sight path and induce airframe shadowing.Reported measurements found the weakest received signal during a right banking turn and the strongest signal during a left banking turn.
A. Airframe Shadowing: Low-altitude multirotor UAV airframe shadowing remains insufficiently characterized because most relevant campaigns used manned aircraft at high altitude.The survey therefore identifies low-altitude multirotor measurements as an open research topic.
A. Airframe Shadowing: Airframe shadowing during circular flight tracks can be modeled by a Gaussian distribution and is more substantial than during linear flight.Other studies attributed shadowing to aircraft wings and engines and modeled shadowing loss and duration empirically.
B. Stationary Interval: UAV channel non-stationarity violates the WSSUS assumption, making stationary-interval estimation important for interpreting wideband frequency-dispersive statistics.A reported C-band measurement estimated the stationary interval at approximately 15 m or 250λ using 50 MHz bandwidth.
C. Diversity Techniques: MIMO measurements reported throughput, range, and SNR improvements, including two-times throughput and 1.6-times range with 4 × 4 MIMO compared with SISO.A 2 × 2 configuration produced throughput gains up to 8 times for most of the flight route, while helicopter-mounted antennas achieved approximately 13 dB SNR gain.

VI. FUTURE RESEARCH CHALLENGES

The paper identifies priorities for UAV channel modeling, including airframe shadowing, non-stationarity, diverse measurement settings, and air-to-air propagation. It also surveys measurement campaigns and modeling approaches to guide future campaigns and practical channel models.

Future measurement campaigns: Measurement campaigns are required to characterize UAV channels in dense urban and metropolitan environments, subject to civil aviation regulatory approval.
Future measurement campaigns: USRP platforms offer flexible, low-power wideband measurements for testing multicarrier and MIMO communication protocols.
Airframe shadowing: Airframe shadowing in small rotary UAVs remains insufficiently studied for both air-to-ground single-hop and air-to-air multihop channels.
Airframe shadowing: Ray tracing can investigate airframe shadowing by incorporating UAV shape, metallic properties, and maneuvering positions through CAD tools.
Channel non-stationarity: Estimating stationary intervals is important for wideband frequency-selective air-to-ground parameters, with spectral divergence and evolutionary methods suggested for this purpose.
Survey contribution: The survey categorizes low-altitude measurement campaigns and UAV channel models, reviews air-to-ground and air-to-air propagation, and highlights challenges involving shadowing, non-stationarity, and diversity.