PaperlayerView exact revision on arXiv

Source-linked AI summary

A Comprehensive Overview on 5G-and-Beyond Networks with UAVs: From Communications to Sensing and Intelligence

Qingqing Wu, Jie Xu, Yong Zeng, Derrick Wing Kwan Ng, Naofal Al-Dhahir, Robert Schober, A. Lee Swindlehurst

arXiv:2010.09317v2cs.ITcs.AIcs.LGcs.NI

TL;DR

Integrating UAVs into 5G-and-beyond networks must support diverse communication, sensing, and intelligence requirements while handling 3D coverage and air-ground interference challenges. This paper surveys research and industry efforts across advanced techniques including IRS, short-packet transmission, energy harvesting, joint communication and radar sensing, and edge intelligence. It organizes current solutions and identifies open directions for UAV-enabled 3D wireless networks.

  • Problem

    UAV integration requires simultaneous support for aerial and ground users, but existing terrestrial techniques may not directly address 3D coverage, air-ground interference, sensing, and intelligence requirements.

  • Method

    The paper provides a comprehensive overview of academic and industrial research on integrating UAVs into cellular networks across communication, sensing, and AI technologies.

  • Results

    The paper synthesizes advanced solutions including IRS, short-packet communication, energy harvesting, joint communication and radar sensing, and edge intelligence, and highlights open research directions.

  • Takeaways & Limitations

    UAV cellular networks can be studied as 5G-and-beyond 3D wireless systems that jointly exploit aerial mobility, advanced wireless technologies, sensing, and intelligence.

Abstract

from arXiv · show

Due to the advancements in cellular technologies and the dense deployment of cellular infrastructure, integrating unmanned aerial vehicles (UAVs) into the fifth-generation (5G) and beyond cellular networks is a promising solution to achieve safe UAV operation as well as enabling diversified applications with mission-specific payload data delivery. In particular, 5G networks need to support three typical usage scenarios, namely, enhanced mobile broadband (eMBB), ultra-reliable low-latency communications (URLLC), and massive machine-type communications (mMTC). On the one hand, UAVs can be leveraged as cost-effective aerial platforms to provide ground users with enhanced communication services by exploiting their high cruising altitude and controllable maneuverability in three-dimensional (3D) space. On the other hand, providing such communication services simultaneously for both UAV and ground users poses new challenges due to the need for ubiquitous 3D signal coverage as well as the strong air-ground network interference. Besides the requirement of high-performance wireless communications, the ability to support effective and efficient sensing as well as network intelligence is also essential for 5G-and-beyond 3D heterogeneous wireless networks with coexisting aerial and ground users. In this paper, we provide a comprehensive overview of the latest research efforts on integrating UAVs into cellular networks, with an emphasis on how to exploit advanced techniques (e.g., intelligent reflecting surface, short packet transmission, energy harvesting, joint communication and radar sensing, and edge intelligence) to meet the diversified service requirements of next-generation wireless systems. Moreover, we highlight important directions for further investigation in future work.

I. INTRODUCTION

UAV integration into 5G-and-beyond networks addresses growing application and connectivity needs while introducing new 3D coverage, interference, sensing, and intelligence challenges.

  • Motivation: The commercial UAV market is projected to grow from 19.3 billion dollars in 2020 to 45.8 billion dollars in 2025, driven by lower costs and expanding applications.Applications include surveillance, aerial imaging, precision agriculture, logistics, law enforcement, disaster response, and emergency care.
  • Motivation: Safe BVLOS operation requires secure, ultra-reliable command-and-control links, while real-time aerial video requires high-rate air-to-ground payload links.Cellular networks can in principle meet these requirements because of their dense deployment and capabilities.
  • 5G Service Scenarios: 5G supports eMBB, URLLC, and mMTC, respectively targeting high data rates, mission-critical low-delay reliability, and massive connectivity for power-limited devices.These three categories define major service requirements for UAV-integrated networks.
  • Challenges and Opportunities: Terrestrial techniques may not directly apply to 3D networks because UAVs and ground users operate in different environments and introduce previously unexplored design degrees of freedom.These degrees of freedom can help enhance communication performance.
  • Challenges and Opportunities: Simultaneously serving aerial and ground users requires elevation-aware beam steering and creates strong air-ground interference challenges.High-altitude UAVs require ground base stations to steer beams in both azimuth and elevation planes.
  • Challenges and Opportunities: UAV mobility can bypass blockage and expand coverage, while sensing supports safe operation and air-traffic management and AI addresses data-intensive, dynamic design problems.The overview considers UAVs as communication platforms and sensing platforms within 3D heterogeneous wireless networks.

B. Industry Progress, Projects, and Standardization

Industry trials, research projects, and 3GPP standardization have advanced UAV cellular networking from experimentation toward integrated support for diverse UAV services.

  • Industry Progress: Industry efforts evaluated UAV cellular connectivity through SINR measurements and studies of power control, resource partitioning, uplink interference, and throughput.Qualcomm, Intel, AT&T, Nokia, Verizon, Ericsson, China Mobile, and Huawei conducted related experiments or deployments.
  • Industry Progress: Nokia’s solar-powered F-Cell addressed backhaul cabling and deployment costs, whereas Huawei’s tethered SkySite provided stable power, secured high-speed backhaul, and unlimited endurance.Both systems used UAV-mounted cellular infrastructure, but their power and backhaul designs differed.
  • Projects: ABSOLUTE used aerial base stations to enhance ground capacity and coverage for terrestrial and satellite communications, especially in public-safety emergencies.The project illustrates UAV platforms serving as temporary or augmenting network infrastructure.
  • Projects: The 5G!Drones project used network slicing to run eMBB, URLLC, and mMTC UAV services on shared 5G infrastructure without affecting one another’s performance.The project involved 20 industrial and academic partners from 8 European countries.
  • Standardization: 3GPP standardization addressed UAV connectivity, identification, and tracking, beginning with enhanced LTE studies for networks using downtilted base-station antennas.The standardization work focused on challenges in applying existing cellular infrastructure to UAVs.
  • Objectives and Contributions: This survey distinguishes itself by jointly covering 3D eMBB, URLLC, mMTC challenges, sensing, AI, and future research directions.It extends earlier overviews that did not cover sensing and AI technologies.

II. ENHANCED MOBILE BROADBAND

The eMBB discussion examines M-MIMO, mmWave, and IRS as tools for high-rate UAV networking, emphasizing IRS as a cost-effective approach to power enhancement and interference suppression.

  • eMBB Requirements: 5G eMBB targets peak data rates of 10 Gbits/s, supporting high-rate UAV applications such as real-time video streaming and data relaying.eMBB also targets moderate reliability, with packet error rates on the order of 10^-3.
  • M-MIMO and mmWave: M-MIMO enables fine-grained 3D beamforming from full-dimensional arrays, helping mitigate interference between high-altitude UAVs and low-altitude terrestrial users.Its beamforming gain depends critically on accurate channel state information.
  • M-MIMO and mmWave: UAV mobility can alleviate mmWave attenuation and blockage by reducing propagation distance or repositioning around obstacles.This makes mobility useful for UAV platforms and users in rate-demanding eMBB scenarios.
  • IRS-Assisted Communications: IRS offers a lower-cost alternative to M-MIMO and mmWave hardware for improving received power and suppressing air-ground interference.Its passive reflecting elements use tunable reflection coefficients to manipulate impinging waves.
  • IRS-Assisted Communications: IRS deployment can reduce UAV access delay and propulsion energy by avoiding sequential visits to IoT devices or clusters.Without IRS, a single UAV may need to approach each device sequentially; multiple UAVs instead increase coordination and signaling overhead.
  • Terrestrial IRS: A 21 dB received power gain was achieved for a 50-meter UAV with a properly deployed 100-element passive IRS under 3GPP ground-to-air channel models.The study also found that optimal IRS altitude decreases as IRS-BS distance increases.

2) UAV IRS Assisted Terrestrial Communications:

UAV-based intelligent reflecting surfaces (UIRSs) offer a cost-effective alternative for terrestrial communications, but their deployment and control remain challenging. Research has examined placement optimization, dynamic positioning, and practical robustness issues.

  • Small UAVs may not carry bulky active RF transceivers, while active relays also increase energy consumption and limit endurance.
  • Under LoS assumptions with a blocked direct BS-user link, UIRS placement is optimized by maximizing the achievable rate or received SNR through horizontal location selection.The setup uses BS-UIRS distance d, UIRS altitude H, and BS-user distance D.
  • Depending on H and D, the optimal UIRS location shifts from the midpoint toward positions equidistant from the BS and user.For D = 400 m, the result is verified using P = 20 dBm, β0 = −30 dB, N = 250, and σ2 = −100 dBm.
  • When D is considerably larger than H, UIRSs should be deployed close to the transceivers, unlike active relays.Practical deployment must also consider link blockage, channel rank, condition number, UIRS scale, cooperation, and element orientation.
  • Existing research has jointly optimized aerial IRS placement, phase shifts, and BS beamforming, while reinforcement learning has supported dynamic UIRS positioning for mobile users.The cited studies include service-area SNR maximization and Q-learning-based deployment for mobile outdoor mmWave users.
  • IRS-assisted eMBB air-ground communications remain at an early research stage, with unresolved problems in robust design and practical deployment.

2) Deployment of IRS/UIRS and UAV-IRS Symbiotic Systems:

The section reviews short-packet communication for URLLC and related UAV relay techniques, emphasizing latency, reliability, and finite-blocklength effects. It also identifies deployment and information-transfer directions for IRS/UIRS systems.

  • 2) Deployment of IRS/UIRS and UAV-IRS Symbiotic Systems:: Future IRS/UIRS research should jointly optimize resource allocation and terrestrial or aerial IRS deployment, including locations and densities.UIRSs also need to transfer their own control and sensor information, motivating symbiotic communication that modulates IRS information onto reflected signals.
  • A. Short-Packet Communication: The total latency comprises transmission, propagation, processing, retransmission, and signaling components that can be reduced through different design approaches.A UAV relay can reduce retransmission time on shadowed links, while non-coherent transmission can reduce processing and signaling latency.
  • 1) Performance Metric for SPC:: Finite-blocklength analysis captures the relationship among achievable rate, decoding error probability, and packet length for UAV communication resource allocation.The normal approximation includes a finite-blocklength penalty and approaches Shannon capacity as blocklength grows.
  • A. Short-Packet Communication: Short packets enable URLLC’s low-latency operation but make decoding errors unavoidable, so Shannon capacity alone cannot accurately characterize performance.Existing UAV resource-allocation algorithms often assume infinite blocklength and provide insufficient flexibility for balancing delay and reliability.
  • 1) Performance Metric for SPC:: A UAV operating as a decode-and-forward relay can establish a two-hop link between a ground base station and user when the direct channel is heavily shadowed.The relay example models the BS, user, and UAV in three-dimensional coordinates with fixed UAV altitude.

2) Example for Resource Allocation for URLLC:

The example jointly optimizes UAV-relay placement and two-hop blocklengths under finite-delay URLLC constraints. Finite blocklength creates a rate gap relative to infinite blocklength, while the best UAV position balances the two hops and depends on transmit powers.

  • System model: The UAV-assisted system models a decode-and-forward relay between a ground BS and user suffering heavy-shadowing direct communication.The BS, user, and UAV occupy specified 3D coordinates, with the UAV held at fixed altitude H.
  • Rate model: The two hop rates depend on UAV position through the received SNRs and are computed using the normal approximation for finite blocklengths.The blocklengths m1 and m2 are assigned to the BS and UAV hops, respectively.
  • Optimization: Resource allocation jointly optimizes UAV position and the two hop blocklengths subject to delay, integer-blocklength, and serving-cell constraints.The resulting problem is non-convex because of the combinatorial blocklength constraint and the normal-approximation objective; penalty and monotonic optimization can solve it optimally, while successive convex approximation gives a polynomial-time suboptimal solution.
  • Evaluation setup: The numerical setup uses a 1 ms maximum delay, 0.01 ms mini-time slots, 200 kHz bandwidth, and at most 100 transmission blocks.The reference-distance CNR is set to 60 dB and the required packet error probability is ϵ = 0.1%.
  • Findings: Finite blocklength achieves a lower rate than infinite blocklength because too few coded information blocks average out Gaussian noise under stringent URLLC delay constraints.The optimal UAV position depends strongly on both BS and UAV power budgets and balances the rates of the two hops.

1) Multiple Access:

UAV-enabled mMTC must address scarce spectrum, demanding ground-air and air-ground resource needs, massive uplink connectivity, reliability, latency, and energy efficiency. UAVs can collect sensor data directly, but battery limits and hardware assumptions constrain practical deployment.

  • Multiple Access: UAV URLLC systems face limited spectrum for simultaneous ground-ground, ground-air, and air-air communications, alongside demanding wireless backhaul and exchange links.These links lack the high-speed fixed-line backhaul typical of terrestrial BSs.
  • Multiple Access: Existing UAV-URLLC algorithms commonly assume perfect hardware, although trajectory, positioning, and radio-chain imperfections can degrade decoding capability.The paper calls for performance analysis that accounts for hardware impairments when assessing URLLC guarantees.
  • Multiple Access: mMTC implementation requires massive connectivity, ultra-high reliability, low latency, and energy-efficient transmission, with especially high uplink resource demand.UAV-mounted cameras exemplify applications that periodically send captured data to ground BSs for analysis.
  • Multiple Access: UAVs can serve as mobile sink nodes for ground sensors, using line-of-sight channels to support low-power transmission and extend sensing-network lifetime.This is particularly relevant for large rural sensor networks with limited cellular coverage.
  • Multiple Access: UAV-enabled mMTC remains constrained by small onboard batteries imposed by size, weight, and power limitations.Sustainable UAV communications therefore require practical methods beyond the benefits of direct collection and relaying.

A. NOMA

NOMA supports simultaneous access for coexisting UAV and ground users, while UAV mobility and cooperative schemes can improve rates under particular traffic and channel conditions. Energy harvesting and wireless power transfer address device-energy constraints, but UAV endurance remains limited by onboard batteries.

  • A. NOMA: NOMA uses superposition coding and successive interference cancellation to enable simultaneous transmissions within one resource block and partially mitigate co-channel interference.It is especially suited to future mMTC applications requiring efficient access for many devices.
  • A. NOMA: NOMA lets UAVs or UAV swarms reuse resource blocks occupied by ground users, increasing support for aerial users under high ground-user density compared with OMA.A reported scheme achieves higher data rates than OMA and noncooperative schemes, especially when ground traffic is congested.
  • A. NOMA: NOMA generally outperforms OMA, but its capacity gain decreases as UAV maximum speed or flight duration increases; at infinitely high speed or long duration, NOMA and TDMA match.The comparison has also been extended to sum-rate and outage-probability objectives under practical Rician air-ground channels.
  • A. NOMA: UAV-NOMA performance depends on user pairing, bandwidth allocation, power allocation, altitude, and antenna beamwidth in multi-user designs.Existing work pairs a cell-centered near user with a cell-edge far user in a multi-user rate max-min problem.
  • A. NOMA: Wireless power transfer and energy harvesting can prolong MTD operation, but harvested energy depends critically on power-source locations.A UAV energy transmitter may hover between widely separated devices to reach the Pareto boundary of their energy region.
  • A. NOMA: UAV-aided WPCNs separate or integrate energy transmission and information reception across UAVs, with optimization targeting system energy or minimum user throughput.Multi-agent deep reinforcement learning has been proposed for max-min optimization in multi-UAV WPCNs.
  • A. NOMA: UAV communications remain endurance-limited because onboard batteries have restricted storage capacity, motivating external power and mechanical dynamic optimization.This constraint applies despite the flexibility and mobility of UAV platforms.

1) External-powered UAV:

External-powered UAV research addresses energy supply, propulsion–communication tradeoffs, and scalable coordination for UAV-enabled services. Radio-based sensing is framed as complementary to onboard sensing, with two paradigms: sensing for UAV and UAV for sensing.

  • External-powered UAV: External-powered UAVs use solar or laser sources, with laser-powered systems charging UAVs in flight for subsequent communication.Laser-assisted designs account for information and energy causality, power, and mobility constraints; trajectory performance depends on wavelength and weather.
  • External-powered UAV: UAV mechanical optimization minimizes communication and propulsion energy while satisfying service objectives such as users’ quality of service.Communication energy includes circuit and transmit-power consumption.
  • External-powered UAV: Zero velocity makes propulsion energy tend toward infinity, whereas hovering can improve service provision, creating a tradeoff between energy consumption and communication throughput.Energy efficiency, defined as throughput divided by UAV energy consumption, is used to balance these goals.
  • External-powered UAV: Energy-efficient UAV design has been studied for applications including UAV-assisted mobile edge computing and backscatter communication.The reviewed works extend energy-efficiency considerations beyond conventional cellular transceiver power models.
  • External-powered UAV: Energy-efficient coordination for many collaborative UAVs remains challenging because cooperation requires excessive data and control exchange.Heuristic and genetic algorithms can be suboptimal or computationally costly, while mean-field game methods model large-scale interactions.
  • External-powered UAV: In wide-area mMTC data collection, bringing UAVs close to every device raises propulsion energy, motivating inter-UAV cooperation and random-access designs for large deployments.Grant-based random access can suffer collision failures and signaling overhead as UAV numbers grow; grant-free access versus NOMA remains open.
  • External-powered UAV: Radio-based sensing complements onboard sensors through two paradigms: sensing for UAV and UAV for sensing.The former supports safe flight and airspace management, while the latter uses UAVs as aerial sensing platforms.

A. Sensing for UAV

Sensing for UAV supports safe flight and detection of potentially hazardous UAVs, while the sensing architecture also models UAV-based radar and its relevant performance parameters. The section emphasizes both operational use cases and sensing-system design considerations.

  • Sensing for UAV: Sense-and-avoid is indispensable for autonomous and semi-autonomous UAV safety, requiring swift collision and obstacle responses.It can also support constant-altitude maintenance for applications such as fertilizer or pesticide spraying.
  • Sensing for UAV: Commercial vision- and light-based sense-and-avoid systems are vulnerable to poor environmental conditions.This motivates radio-based alternatives and complementary sensing approaches.
  • Sensing for UAV: Hazardous or illegitimate UAVs require detection, tracking, and classification because they may be non-cooperative or deceptive.Passive radar based on echoed or scattered signals is needed when active detection and localization are infeasible, while small radar cross sections make UAV detection difficult.
  • Sensing for UAV: Radio-based UAV sensing research includes radar detection, micro-drone classification, UAV-based sensing platforms, and synthetic-aperture radar for higher angular resolution.The reviewed literature is organized in a comparison table covering radio-based UAV sensing works.
  • Sensing for UAV: UAV sensing benefits from wider field of view, reduced blockage, and controllable 3D mobility that enables trajectory optimization for target tracking.Potential applications include law enforcement, precision agriculture, mapping, search and rescue, and military operations.
  • Sensing for UAV: The UAV radar input-output model represents reflected target signals using transmit and receive array responses, propagation delays, target angles, coefficients, and receiver noise.For UAV sensing, both azimuth and elevation angles matter because of the elevated platform; the model can apply to bi-static and mono-static radar.
  • Sensing for UAV: Radar sensing detects target presence or absence and estimates parameters such as angle and delay to determine target state, using detection probabilities, false alarms, accuracy, MSE, or CRLB-related measures.Target delay is related to distance and radial velocity, and MIMO radar can improve effective angular resolution through virtual-array elements.
  • Sensing for UAV: MmWave sensing makes small objects such as micro-UAVs more electrically visible than in sub-6 GHz systems because of its shorter wavelength.Radar cross section is used to illustrate this detectability advantage.

D. Joint UAV Communication and Sensing

Joint communication and sensing evolves from coexistence with interference management toward dual-function systems that implement both functions in one device. UAV networks motivate this integration because sensing and communication may operate concurrently in shared environments.

  • D. Joint UAV Communication and Sensing: Communication-radar coexistence keeps separate waveforms and transceiver designs while controlling mutual interference through resource allocation.A low-altitude UAV broadcasting radar waveforms must account for interference with nearby base stations and ground users.
  • D. Joint UAV Communication and Sensing: Dual-function radar-communications implements radar and communication functions in the same device, enabling a cellular base station to serve users while monitoring low-altitude airspace.DFRC designs may be radar-centric, communication-centric, or integrated.
  • D. Joint UAV Communication and Sensing: Radar-centric DFRC embeds information-bearing symbols into radar waveforms or beam patterns but supports only very low data rates despite simple, power-efficient implementation.Examples include phase or frequency modulation of FMCW pulses and amplitude modulation of radar sidelobes.

E. Future Research

Future research must address hardware, endurance, heterogeneity, and the joint control of UAV mobility, communications, computation, and intelligence. The overview identifies reinforcement learning and distributed edge learning as important directions while noting unresolved scalability and modeling challenges.

  • E. Future Research: UAV sensing remains constrained by SWaP requirements, demanding compact, lightweight, and energy-efficient sensing devices.Limited endurance also makes all-day aerial sensing networks difficult to build using UAVs alone.
  • E. Future Research: Future sensing networks are expected to combine ground, aerial, and space platforms to exploit complementary sensing capabilities.This heterogeneous architecture is proposed to maximize sensing performance.
  • E. Future Research: AI can support UAV trajectory and communication design when accurate models are difficult, while UAV edge intelligence must handle computation-intensive and latency-critical tasks.Joint mobility, communication, and computation allocation is difficult when UAVs act as aerial users, edge servers, or both.
  • E. Future Research: The literature lacks an overview of AI integration in UAV networks covering learning-based design and edge intelligence across the identified aspects.The section reviews trajectory and communication learning, UAV MEC offloading, distributed edge machine learning, and open problems.
  • E. Future Research: Reinforcement learning is useful for joint UAV movement and communication design because it optimizes agent actions from cumulative rewards over time.It is especially relevant when channel LoS/NLoS information is uncertain and model-driven optimization is infeasible.
  • E. Future Research: For a single UAV acting as an aerial base station, model-free reinforcement learning can learn trajectory solutions from experience without prior LoS/NLoS information.Environment maps can additionally provide spatial channel information for more efficient trajectory design under obstacle-avoidance constraints.
  • E. Future Research: Multiple UAVs require collaborative trajectory and communication design to improve coverage and throughput for ground users.Existing approaches include K-means, Q-learning, and decentralized deep reinforcement learning.

2) UAVs as Users:

UAVs acting as cellular users must maintain connectivity while completing missions, but blockage, interference, and coupled computation offloading make trajectory and resource design challenging.

  • Cellular-connected UAVs perform tasks such as packet delivery and aerial inspection while maintaining their cellular connection.
  • Conventional distance-based trajectory designs can overlook air-ground propagation complexity and cause outages in coverage holes created by blocked ground-BS channels.
  • Co-channel interference among nearby base stations further affects performance, while altitude changes the line-of-sight probability and channel path loss.
  • Computation Offloading: Computation offloading requires jointly designing UAV trajectories, offloading schedules, and downloads because results must return through the same base station.This coupling can force the UAV to fly back and forth during offloading and downloading.
  • Computation Offloading: Research extends computation offloading to parallel execution across base stations, UAV-mounted MEC servers, relay-assisted MEC, multiple UAV servers, and swarm-support architectures.These setups introduce coordination, association, communication, computation, and resource-allocation requirements.

C. Distributed Edge Machine Learning with UAVs

Distributed edge learning lets UAVs collaboratively train models from local data, but UAV-enabled learning remains immature and must support dynamic aerial-ground resource matching.

  • Federated edge learning enables UAVs to train common models from distributed data without sharing the data, supporting security and privacy.
  • UAV-enabled federated learning research has examined wireless resource allocation, UAV communication applications, and UAV swarms, but remains in its infancy.
  • Future aerial-ground networks must match time- and spatial-varying communication, computation, and AI demands with distributed heterogeneous resources.
  • Federated edge learning architectures may coordinate with ground base stations or use UAV peers as shown in the two cases of Fig. 9.
  • AirComp integrates communication and computation through wireless signal superposition, with applications in distributed sensing, federated learning, and UAV swarms.
  • The overview reviews UAV integration into 5G-and-beyond networks and identifies open problems in communication, sensing, AI, and 3D wireless-network design.
Loading 2010.09317v2…