PaperlayerView exact revision on arXiv

Source-linked AI summary

Five Facets of 6G: Research Challenges and Opportunities

Li-Hsiang Shen, Kai-Ten Feng, Lajos Hanzo

arXiv:2212.07902v1cs.NIeess.SY

TL;DR

The paper addresses pivotal issues in emerging 6G scenarios by surveying promising techniques and organizing performance metrics and service use cases. It reviews hybrid architectures and learning-driven schemes, and advocates multi-component Pareto optimization for complex network requirements.

  • Problem

    Emerging 6G scenarios present pivotal issues, including cloud storage and requirements for vertical and horizontal massive ultra-dense networks.

  • Method

    The paper surveys promising 6G techniques, organizes key performance metrics and service use cases, and considers machine and deep learning schemes for different requirements.

  • Results

    The survey describes hybrid networks delivering smooth, resilient, high-quality services and reports a lowest localization error of 0.15 meter for a proposed method.

  • Takeaways & Limitations

    The paper advocates moving from single-component toward multi-component Pareto optimization for future complex network scenarios with differing requirements.

Abstract

from arXiv · show

Whilst the fifth-generation (5G) systems are being rolled out across the globe, researchers have turned their attention to the exploration of radical next-generation solutions. At this early evolutionary stage we survey five main research facets of this field, namely {\em Facet~1: next-generation architectures, spectrum and services, Facet~2: next-generation networking, Facet~3: Internet of Things (IoT), Facet~4: wireless positioning and sensing, as well as Facet~5: applications of deep learning in 6G networks.} In this paper, we have provided a critical appraisal of the literature of promising techniques ranging from the associated architectures, networking, applications as well as designs. We have portrayed a plethora of heterogeneous architectures relying on cooperative hybrid networks supported by diverse access and transmission mechanisms. The vulnerabilities of these techniques are also addressed and carefully considered for highlighting the most of promising future research directions. Additionally, we have listed a rich suite of learning-driven optimization techniques. We conclude by observing the evolutionary paradigm-shift that has taken place from pure single-component bandwidth-efficiency, power-efficiency or delay-optimization towards multi-component designs, as exemplified by the twin-component ultra-reliable low-latency mode of the 5G system. We advocate a further evolutionary step towards multi-component Pareto optimization, which requires the exploration of the entire Pareto front of all optiomal solutions, where none of the components of the objective function may be improved without degrading at least one of the other components.

I. INTRODUCTION

The introduction frames 6G as a transition toward heterogeneous hybrid networks with stringent service demands, new architectural and networking challenges, and security concerns. It surveys five research facets, reviews promising techniques and vulnerabilities, and advocates multi-component Pareto optimization.

  • Next-generation architectures: Future 6G networks are expected to extend beyond conventional terrestrial cellular architectures toward hybrid terrestrial-underwater-aerial-space systems.The introduction associates this evolution with heterogeneous network architectures and diverse access and transmission mechanisms.
  • Motivation: 6G research targets stringent potential specifications, including a peak rate of 1 Tbps, latency below 0.1 ms, and reliability of 99.99999999%.The paper notes that these performance indicators were not formally finalized and were drawn from open literature.
  • Survey scope: The survey organizes open issues into five facets: architectures, spectrum and services; networking; IoT; wireless positioning and sensing; and deep learning applications.These facets span architectures, networking, applications, and scheme designs extending current wireless and networking foundations.
  • Contributions: The paper reviews heterogeneous architectures, vulnerabilities, learning-driven optimization schemes, and field-trial results involving high-frequency communications, IoT, sensing, and device-free positioning.It presents the review as a cross-disciplinary synthesis covering network layers, architectures, applications, and optimization.
  • Optimization perspective: The paper advocates moving from single-component optimization toward multi-component Pareto optimization over trade-offs among rate, bandwidth, energy, latency, and complexity.This principle seeks solutions where improving one objective requires degrading at least one other objective.

A. Advanced Wireless Network Architecture and Technology

The section surveys heterogeneous 6G architectures spanning terrestrial, aerial, satellite, underwater, and user-centric networks, alongside high-frequency links and flexible CU/DU/RU deployments. These designs broaden coverage and capacity but introduce substantial challenges in propagation, mobility, interference, synchronization, and resource management.

  • Heterogeneous architectures: 6G architectures extend beyond conventional terrestrial networks to integrate underwater, aerial, satellite, and heterogeneous ground-air-space communications.These architectures support applications including autonomous underwater vehicles, aerial platforms, satellites, and cooperative information transfer.
  • Flexible network architecture: Flexible CU/DU/RU architectures distribute computation, storage, lower-layer functions, signal transmission, and networking policy across cloud and edge units.Their fronthaul and backhaul may use high-speed mmWave and THz links instead of traditional wired connections.
  • Non-terrestrial and underwater networks: Underwater, aerial, and satellite networks improve specialized connectivity but remain constrained by environmental disturbances, long-distance pathloss, mobility, and power requirements.Underwater links face water flow, Doppler effects, noise, and vibration; satellite systems require high transmit power and specialized terrestrial-space terminals.
  • User-centric networking: User-centric cell-free networks allocate resources to UE QoS requirements, creating amorphous coverage areas and strong load-balancing capability.Realizing this design across diverse cell sizes still requires further research, alongside interference management, channel allocation, handover, admission control, power control, and scheduling.

B. New Multiuser Transmission Schemes

The section reviews multiuser transmission schemes including NOMA, 3D beamforming, full duplex, CoMP, RSMA, and RIS-assisted designs for diverse 6G resource domains. These techniques offer spectral, throughput, coverage, or interference-management benefits while leaving substantial challenges in coordination, beamforming, and resource optimization.

  • NOMA: NOMA multiplexes multiple signals in shared resource slots and uses successive interference cancellation to separate users.Its integration across mmWave/THz resource domains, interference cancellation in ultra-dense networks, and coverage enhancement for UAVs and satellites remain challenging.
  • 3D beamforming: 3D beamforming serves users at different angles, while high-gain beams are critical for compensating mmWave and THz pathloss.The design must manage limited three-dimensional resources while supporting diverse QoS requirements.
  • Full duplex: Full duplex enables simultaneous uplink and downlink transmission at the same frequency, but high-power self-interference requires advanced mitigation.mmWave/THz ultra-massive MIMO beamforming offers an opportunity to increase spectral efficiency in terrestrial and aerial networks.
  • CoMP: CoMP coordinates simultaneous transmissions from multiple base stations and, with mmWave/THz beamforming, can increase network throughput.Large-scale deployment faces synchronization difficulties because a receiver can be perfectly synchronized with only one base station.
  • RSMA: RSMA partitions each user’s message into common and private segments, jointly transmitting them and using SIC to recover the private signals.The described formulation relies on simplifying assumptions, while users treat other private signals as noise because their precoding weights are unknown.
  • RIS-assisted transmission: RISs use many phase-adjustable metamaterial elements to reflect waves, potentially extending coverage, reducing power consumption, and enhancing data rates.RIS-assisted designs require joint active beamforming and passive phase-shift optimization because RIS deployment can also create additional interference.
  • RIS-assisted transmission: RIS-NOMA, RIS-FD, and related RIS architectures target channel separation, self-interference reduction, and coverage extension across multiuser transmissions.Jointly designing RIS and NOMA across time, frequency, and spatial domains remains a wide open research issue.

C. Unlicensed Spectrum Access

Unlicensed-spectrum access in 6G spans sub-6 GHz, mmWave, and THz technologies, combining wider bandwidth with beamforming and coordinated multiuser transmission. Key challenges include interference, propagation loss, beam-training overhead, and cross-spectrum coordination.

  • mmWave Frequencies: Coordination-based beamforming training targets near-unity user association and tolerable beam-alignment outage while allowing APs to tune training and contention durations.Users perform association and individual beam training, whereas APs adjust training-frame and contention-slot lengths.
  • mmWave Frequencies: The CBFT scheme imposes the lowest latency and achieves the highest throughput among the compared time-division and conventional 802.11ad/ay methods.This comparison concerns coordinated multiuser beamforming training in multi-AP mmWave scenarios.
  • THz Frequencies: THz systems offer abundant bandwidth and up to 100 Gbps-level mid-range fronthaul or backhaul, but require new low-complexity beam training and robust hardware designs.Challenges include high-power signal generation and detection, hostile propagation, higher beam resolution, and low-overhead coordination across hybrid-radio networks.
  • THz Frequencies: Cross-spectrum hybrid access remains open because coordinative beam training and transmission must operate under limited computing and communication resources.The envisioned architecture combines mmWave wide-beams with THz pencil-beams in collaborative transmission.

D. Multiple Wireless Services

6G services extend beyond 5G eMBB, URLLC, and mMTC toward higher-rate, ultra-reliable, low-latency, long-distance, high-mobility, and extremely low-power communications. This expansion requires flexible resource allocation, hybrid numerology, and integrated heterogeneous networks.

  • Service Evolution: Beyond conventional 5G services, researchers discuss long-distance and high-mobility communications and extremely low-power communications.These service concepts are presented alongside eMBB, URLLC, and mMTC in the 5G-to-6G service landscape.
  • Service Evolution: Next-generation eMBB services require new transmission and access technologies to achieve higher data rates in new wireless architectures.This development is characterized as the enhanced ultra-mobile broadband philosophy.
  • Service Evolution: Enhanced URLLC targets instant, highly reliable, low-latency communications for unmanned factories, aircraft, vehicles, and intelligent transportation systems.These applications require both high reliability and low latency.
  • Service Evolution: Ultra-mMTC is motivated by increasing connection densities and seeks more flexible, efficient, low-latency, and highly adaptive protocols.Hybrid services such as URLLC-eMBB also create open issues in front-end resource allocation and hybrid numerology optimization.
  • 3GPP Evolution: 6G-era heterogeneous radios are tentatively associated with powerful AI techniques and new transmission and spectrum technologies from Release 21 onward.Earlier releases emphasize terrestrial use cases and advanced 5G architectures and spectrum utilization.
  • 3GPP Evolution: Integrating networks horizontally and vertically is presented as necessary to support high transmission rates, remote-area coverage, high mobility, and lower-power IoT devices.The integration goal spans heterogeneous network architectures and service requirements.

A. Network Softwarization of SDN/NFV

SDN and NFV underpin 6G network softwarization by enabling automated management, virtualized functions, flexible deployment, and network slicing. Open issues include security and packet processing across switches in SDN/NFV-enabled cores.

  • SDN/NFV Functions: SDN research addresses automated network management and traffic optimization through routing, load balancing, multipath routing, QoS management, and route repair.These functions support dynamic real-time management of network traffic and failures.
  • SDN/NFV Functions: NFV partitions physical infrastructure into independent virtual networks with separate operating resources and heterogeneous service qualities.This virtualization model supports differentiated services across shared infrastructure.
  • SDN/NFV Functions: NFV increases deployment flexibility while reducing device costs and operating costs, but controller attacks and service blocking remain security challenges.The passage identifies both operational benefits and security vulnerabilities of NFV.
  • SDN/NFV Functions: NFV-MANO combined with SDN principles and optimized traffic-route management is identified as a pivotal research direction for network softwarization.The emphasis is on management and orchestration for software-defined, virtualized networks.
  • Core-Network Architecture: Integrating SDN and NFV supports flexible telecom applications and network slicing, which partitions a physical network into multiple QoS-guaranteed virtual networks.The integration also motivates virtualized evolved packet-core designs.
  • Core-Network Architecture: Flexible packet processing across switches remains a salient research issue in future SDN/NFV-enabled 6G core networks.The challenge concerns packet handling across the control and data layers and different switches.

C. Next-Generation Mobile Network Architecture and Management

Next-generation mobile networks combine cloud, edge, fog, SON, and heterogeneous IoT architectures to improve flexibility, automation, latency, and service quality. Their design must coordinate diverse protocols and resources while balancing efficiency, reliability, integration, power, and security requirements.

  • Mobile Network Architecture: Next-generation mobile networks emphasize flexibility, intelligence, automation, and advanced mobile cloud and edge computing.Adaptive assignment of computing resources between cloud and edge is identified as a crucial task.
  • Mobile Network Architecture: MEC and fog computing are presented as important architectures that can potentially reduce service latency and improve spectral efficiency and QoS.These architectures distribute computing beyond a fully centralized model.
  • Mobile Network Management: As heterogeneous networks add more base stations, manual power allocation and BS deployment become infeasible, motivating self-organized network concepts incorporating SDN and NFV.SON supports self-configuration, self-optimization, self-healing, and self-sustenance, including automatic parameter optimization.
  • IoT Architectures: IoT technologies support diverse configurations, including large-scale access, long-distance transmission, low-power or low-rate operation, and low deployment cost.Examples include Zigbee, Z-Wave, BLE, Bluetooth 5.0, SigFox, LoRa, NB-IoT, and WiFi HaLow.
  • IoT Challenges: IoT data collection requires efficient environmental sensing and uploading, but protocol coordination must also address power consumption, capacity, and spectral efficiency.Traffic management, resource allocation, encryption, and tradeoffs among bandwidth, power, synchronization, and processing complexity remain challenges.
  • IoT Challenges: Low-power IoT design must balance narrow bandwidth, low power consumption, tight synchronization, and limited processing complexity.These requirements arise because many existing IoT technologies rely primarily on narrowband communications.
  • IoT Architectures: IoT research must coordinate heterogeneous architectures and technologies across sensing, networking, transport, operations, management, and application layers while improving efficiency, reliability, and integration.The scope includes personal wearables, industrial IoT, intelligent homes, and underwater Internet applications.

B. Vehicular Networks

Vehicular networks support diverse V2X services and IoT-enabled sensing, but remain constrained by mobility, heterogeneity, routing, security, privacy, and joint sensing-control-communication requirements.

  • V2X services: Connected autonomous vehicles use onboard IoT devices and sensors to detect and track pedestrians, vehicles, traffic signs, and road conditions.V2X can provide long-range detection of hazards and traffic conditions through connected sensors and roadside units.
  • V2X services: V2X supports information exchange among vehicles and connected devices, including V2I, V2V, V2N, V2P, and V2D services.Cellular C-V2X and NR-V2X use cellular networks, while DSRC uses WLAN technology for direct V2V and V2I communication.
  • Open challenges: IoT-V2X research challenges include wireless-channel characteristics, resource management, heterogeneous interfaces, dynamic topology, routing, congestion, security, reliability, and joint sensing, control, and communications.The cited challenges span both network operation and integrated sensing-control design.
  • Social IoT: Social IoT networks connect geographically adjacent devices and users through proximity, D2D, or vehicular communications to support interaction and services.Potential applications include advertising, geographic or content sharing, robotic systems, gaming, and relaying services.
  • Security and privacy: Security and privacy remain open issues across wearables, smart grids, vehicular networks, cloud, and edge systems, while existing protections often address only a single functionality or service.Vehicular concerns include emergency messages, authentication, and ultra-low-latency communication under high mobility.
  • Positioning: Outdoor positioning is limited by GPS accuracy, coverage, signal loss, and high-speed Doppler shift, motivating cellular signals and AI-assisted tracking for vehicles and UAVs.The paper also identifies long-distance or low-power IoT integration with cellular vehicle tracking as a promising service direction.

B. Indoor Positioning

Indoor positioning is moving from device-based fingerprinting toward device-free, AI-assisted sensing that uses wireless signals and distributed edge-central processing, but accuracy depends on stable signals and substantial databases.

  • Positioning modalities: GPS provides almost no indoor coverage, motivating systems based on WiFi, infrared, ultrasonic, visible light, and sensors such as accelerometers, gyroscopes, and magnetometers.CSI is used for centimeter-level positioning because it provides frequency, power, and latency information beyond received signal strength.
  • Applications and challenges: Positioning performance depends on wireless-signal stability and large databases, whose construction is extremely laborious.Spatial skeleton databases inferred from indoor maps are suggested as one mitigation.
  • Device-free positioning: Device-free indoor positioning operates without wearable devices and uses multiple access points with AI-assisted edge and central servers.The architecture collects wireless CSI between AP pairs, trains local models at edge servers, and merges them into a global model centrally.
  • Fingerprinting: Front-end access points measure received-signal fluctuations to build fingerprints representing indoor motion, positioning, presence, and vitality.These fingerprints support databases for indoor activity characterization.
  • Applications and challenges: Device-free monitoring can support continuous tracking without revealing user identities and is especially suitable where wearable devices are unavailable.The paper identifies hostile environments such as oil tankers, mining pits, and complex plants as difficult positioning settings.

C. Wireless Indoor Detection

Wireless indoor detection covers presence, motion, vitality, and pedestrian tracking through signal variations, while facing environmental interference, multipath, limited range, and multi-object complexity.

  • Detection challenges: Multi-object pedestrian tracking remains an open issue because overlapping time- and frequency-domain signals complicate trajectory estimation.The paper also calls for improved physical-layer processing that raises device-free precision at reduced computational complexity.
  • Detection tasks: Wireless indoor detection targets presence, motion, vitality, and pedestrian trajectories using variations in received signals.Applications include smart homes, green buildings, factory manufacturing, healthcare, and pedestrian path tracking.
  • Detection challenges: Presence detection is vulnerable to humidity, temperature, interfering objects, attenuation, multipath, and cross-room variation, requiring database replenishment.Expanding coverage requires balancing deployment cost, implementation complexity, and detection accuracy.
  • Detection challenges: Motion detection can distinguish behaviors from wireless-path changes, but interference from other objects can severely reduce accuracy.The paper lists standing, hand-waving, falling, slow walking, and jumping as example behaviors.
  • Detection challenges: Vitality detection analyzes subtle wireless-signal changes such as breathing and heart-rate variation to identify conditions including arrhythmia and apnea.Fine-grained detection methods remain confined to small areas or short distances, and wearable devices can be inconvenient and require frequent battery recharging.
  • CSI-based results: A device-free WiFi CSI positioning scheme achieves a lowest localization error of 0.15 meter compared with existing positioning methods.The cited platform also evaluates presence and vitality detection using CSI.
  • CSI-based results: The cited CSI-based presence framework has the highest F1-based detection accuracy, while device-free breath detection may approach wearable vitality-detection performance.The experiments include presence observations and abnormal, apnea, and normal breathing scenarios.

VI. FACET 5: APPLICATIONS OF DEEP LEARNING IN 6G NETWORKS

The paper surveys deep-learning applications for 6G communication and networking, covering supervised, unsupervised, reinforcement, transfer, distributed, federated, and related learning approaches. It maps these methods to wireless optimization, network management, IoT, positioning, and sensing while identifying unresolved data, scalability, confidence, and convergence challenges.

  • Scope and motivation: Deep learning is presented as a way to address nonlinear and non-convex 6G communication and networking problems across vertical and horizontal networks.The surveyed mechanisms include supervised, unsupervised, reinforcement, distributed, federated, and transfer learning.
  • Learning mechanisms: Supervised learning uses ground-truth labels, but practical deployment remains challenged by data collection, analytics, and laborious labeling.Beamforming measurements or convex-optimization solutions can provide training labels for neural-network models.
  • Learning mechanisms: Unsupervised learning infers hidden features from unlabelled data for partitioning, clustering, augmentation, security, and large-scale processing.The surveyed tools include GANs, PCA, HMM, and EM for information insufficiency, dimensionality, and uncertain environments.
  • Learning mechanisms: Reinforcement learning supports dynamic policy adaptation, resource management, scheduling, and access, but faces convergence-speed, uncertainty, optimality, and theoretical-proof challenges.DQN extends Q learning with neural networks, while DDPG uses action and critic networks for policy selection and evaluation.
  • Learning mechanisms: Transfer learning reuses models trained in one network or environment to improve retraining efficiency in new environments.The paper relates this approach to complex wireless propagation and high tele-traffic requirements.
  • Network learning architectures: Distributed learning parallelizes tasks across computing units, while intelligent 6G core units can separately manage control and user planes and dynamically configure network functions.This addresses overload associated with fully centralized computing.
  • 6G applications: Meta learning adapts trained parameters to heterogeneous tasks from sparse samples and labels, with reported use in cellular, IoT, and vehicular networks.The cited description emphasizes few experiences, fast convergence, and low computational complexity.
  • 6G applications: AI models can process massive IoT data, detect networks for interference avoidance, improve spectrum-versus-energy efficiency, maintain QoS, and adaptively collect environmental data.The paper also associates deep neural networks with support for numerous IoT devices.

VII. SUMMARY AND THE ROAD TO MULTI-COMPONENT PARETO-OPTIMIZATION

The paper surveys five next-generation wireless research topics and frames 6G design as a shift from single-component optimization toward multi-component Pareto optimization. It argues that Pareto fronts should capture all optimal configurations across competing objectives, while learning-driven methods can support such designs.

  • Research scope: The survey covers next-generation architectures, spectrum and services, networking, IoT, wireless positioning and sensing, and deep-learning applications in 6G.
  • Research scope: The paper reviews promising architectures, networking approaches, applications, and scheme designs while identifying open research issues for next-generation wireless.
  • From single- to multi-component optimization: 5G’s simultaneous low-latency and low-bit-error-rate requirements motivate a shift from single-component objectives toward multi-component optimization.
  • Pareto optimization: Unlike finite-block-length information theory, which quantifies performance for a coding length while remaining oblivious to block-code complexity, Pareto optimization includes delay and complexity together.
  • Pareto optimization: A Pareto front contains optimal configurations in which improving any objective, such as throughput, delay, power, or bit error rate, degrades at least one other objective.
  • Learning-driven optimization: Learning-driven approaches include multi-task learning, weighted loss functions, federated learning, reinforcement learning, and transfer learning for diverse optimization requirements.
Loading 2212.07902v1…