G-Advanced, 6G and Emerging Network Paradigms

Introduction and Context

Mobile communication has evolved from voice-centric cellular systems to programmable, cloud-based and increasingly intelligent networks. Fifth-generation (5G) systems introduced a service-oriented 5G Core, massive multiple-input multiple-output (MIMO), millimetre-wave access, ultra-reliable low-latency communication (URLLC), network slicing and private cellular networks. These capabilities support industrial automation, connected vehicles, immersive media, massive sensing and critical infrastructure.

The next stage is not simply a faster radio interface. Advanced research associated with 5G-Advanced and 6G concerns a convergence of communications, sensing, artificial intelligence (AI), distributed computing, non-terrestrial networks, high-frequency transmission and potentially quantum technologies. The central design question is how networks can provide dependable, secure and context-aware services while operating across terrestrial, aerial, satellite and highly heterogeneous devices.

The term “6G” currently denotes a research and standardisation direction rather than a universally finalised technical specification. Requirements, spectrum policies and architectures remain under development. Consequently, this chapter distinguishes deployed or standardised 5G capabilities from forward-looking research proposals.

Learning Objectives

After studying this chapter, the reader should be able to:

1. Explain the architecture and service logic of the 5G Core.

2. Analyse the principles and trade-offs of massive MIMO and URLLC.

3. Evaluate network slicing, private networks and Open RAN.

4. Explain semantic and AI-native communication models.

5. Describe integrated sensing and communication, terahertz links and quantum-networking directions.

6. Assess technical, economic, security, ethical and regulatory implications of emerging network paradigms.

1. 5G-Advanced as a Transition Platform

5G-Advanced may be understood as the evolution of 5G radio and core capabilities through later releases, enhanced performance, improved positioning, reduced energy consumption, extended industrial support and greater use of AI. Important themes include enhanced uplink, improved mobility, more capable positioning, reduced-capability devices, non-terrestrial integration and tighter support for industrial time-sensitive applications.

A useful conceptual model is the network capability stack:

• Connectivity: radio access, transport and core networking.
• Programmability: APIs, cloud-native functions and policy control.
• Service assurance: latency, reliability, availability and isolation.
• Intelligence: prediction, optimisation, automation and semantic processing.
• Environment awareness: sensing, localisation and context extraction.

This stack reveals why 6G cannot be judged by peak data rate alone. A network may deliver high throughput but still fail an industrial control application if its tail latency, synchronisation or availability is inadequate.

2. The 5G Core and Service-Based Architecture

The 5G Core (5GC) replaces the more tightly coupled architecture of earlier generations with a service-based architecture (SBA). Network functions expose services through controlled interfaces, generally using service discovery and communication mechanisms suitable for cloud deployment. Important functions include:

• AMF (Access and Mobility Management Function): registration, connection management and mobility procedures.
• SMF (Session Management Function): protocol data unit (PDU) session establishment, IP addressing and user-plane selection.
• UPF (User Plane Function): packet forwarding, traffic steering, quality-of-service enforcement and local breakout.
• AUSF and UDM: authentication and subscriber-data functions.
• PCF (Policy Control Function): policy and quality-of-service decisions.
• NRF (Network Repository Function): discovery of available network functions.
• NSSF (Network Slice Selection Function): selection of an appropriate network slice.
• NEF (Network Exposure Function): controlled exposure of network capabilities to applications.
• AF (Application Function): application-level interaction with policy and traffic management.

The SBA supports modularity and independent scaling. For example, a UPF can be placed near a factory to reduce transport delay, while centralised functions continue to serve authentication and policy requirements. This is a principal enabler of multi-access edge computing (MEC).

Sessions, Quality of Service and Network Exposure

A device can maintain one or more PDU sessions, each associated with a data network and a set of quality-of-service flows. A flow is identified and treated according to policy, including priority, guaranteed or non-guaranteed bit rate and an associated 5G QoS Identifier (5QI). The architecture separates control-plane decisions from user-plane forwarding, allowing traffic to be locally processed without moving all packets through a distant core.

The 5GC also supports exposure of selected capabilities to authorised applications. Exposure must be governed carefully: an API that reveals location, device status or traffic characteristics can be useful for industrial control but can also create privacy and security risks.

Architectural Limitations

Cloud-native modularity introduces operational complexity. Service discovery, certificate management, orchestration, state consistency and observability become critical. A distributed core also enlarges the attack surface. Virtualisation does not automatically provide isolation; isolation depends on software configuration, hardware trust, identity management and continuous monitoring.

3. Massive MIMO and Advanced Radio Access

Massive MIMO uses a base station with many antenna elements to serve multiple users through spatial processing. In a simplified narrowband uplink model,

\[

\mathbf{y}=\mathbf{H}\mathbf{x}+\mathbf{n},

\]

where \(\mathbf{x}\) is the transmitted user vector, \(\mathbf{H}\) is the channel matrix, \(\mathbf{y}\) is the received vector and \(\mathbf{n}\) is receiver noise. A linear detector such as maximum-ratio combining uses the channel estimate to align and combine received signals; zero-forcing additionally attempts to suppress inter-user interference.

Under an idealised single-user, independent, coherent-combining model, received signal power can increase approximately with the square of the number of coherently combined antenna elements, while practical multi-user gains depend on channel correlation, pilot quality, propagation and hardware constraints. This ideal scaling must not be interpreted as a universal law.

Critical Design Issues

1. Channel estimation: The transmitter and receiver require sufficiently accurate channel state information. Pilot contamination, pilot overhead and channel ageing limit performance.

2. Spatial correlation: Closely spaced or similarly illuminated antennas may provide less independent spatial information.

3. Energy and hardware cost: Radio-frequency chains, converters, calibration and cooling can offset spectral-efficiency gains.

4. Beam management: High-frequency systems require beam training and tracking, especially under mobility and blockage.

5. Near-field operation: Very large apertures can create spherical-wave rather than plane-wave propagation, requiring new channel models and focusing methods.

Massive MIMO is therefore an optimisation problem involving rate, coverage, energy, hardware complexity and reliability rather than merely an antenna-count target.

4. URLLC: Reliability, Latency and the Tail of the Distribution

URLLC targets applications for which late or lost packets may be more damaging than average throughput reduction. Examples include motion control, protection systems and selected vehicular functions. Latency must be defined precisely: radio-interface latency, transport latency, processing latency and application-to-application latency are different quantities.

Reliability is likewise a probability statement. If \(P_e\) denotes the probability that a packet fails to meet its delivery requirement under specified channel and system conditions, reliability is often represented as \(1-P_e\). The statement is meaningful only when packet size, deadline, channel conditions, traffic load, device population and observation interval are specified.

A useful latency decomposition is

\[

T_{end}=T_{UL}+T_{proc}+T_{transport}+T_{queue}+T_{DL},

\]

where the terms represent uplink transmission, processing, transport, queueing and downlink components. The equation is an accounting decomposition, not a guarantee. A system meets an end-to-end requirement only if the combined distribution, including tail behaviour, satisfies the target.

URLLC techniques include short transmission time intervals, configured or grant-free access, packet duplication, diversity, edge computing, pre-emption and robust scheduling. These techniques impose costs: short packets have coding inefficiency, duplication consumes resources, pre-emption can harm enhanced mobile broadband, and edge sites increase deployment and security requirements.

For critical control, deterministic networking, accurate clock synchronisation, redundant paths, local fallback control and formal safety analysis may be as important as radio reliability. Wireless communication alone cannot guarantee the safety of a physical process.

5. Network Slicing

Network slicing creates logically differentiated networks over shared physical and virtual infrastructure. A slice may be designed for broadband access, industrial control, public safety or massive sensing. Its design includes radio resources, transport, core functions, computing placement, policy and assurance mechanisms.

The lifecycle typically includes design, admission, instantiation, configuration, monitoring, scaling and termination. Slice selection may consider subscriber identity, application intent, location, device capability and service-level agreements.

Slicing is not equivalent to simple virtual LAN separation. Meaningful isolation may require dedicated capacity, separate security domains, independent policy, resource quotas and fault containment. Shared components can create common-mode failures. Slice assurance should monitor throughput, latency percentiles, loss, availability, energy and security events rather than only average rate.

A major research issue is cross-domain slicing. A service may traverse a radio network, several transport domains, edge clouds and a public cloud. End-to-end assurance then requires common semantics for service-level objectives and trusted evidence from each administrative domain.

6. Private 5G and Industrial Networks

A private mobile network is deployed for a defined organisation, site or community of users. It may use locally managed spectrum, shared spectrum or an arrangement with a public operator. Typical components include an on-site radio access network, local core or UPF, identity systems, edge computing and industrial gateways.

Advantages include controlled coverage, predictable mobility, local data handling, integration with operational technology and policy tailored to the site. Limitations include spectrum licensing, radio planning, device ecosystem compatibility, integration with legacy protocols, maintenance capability and total cost of ownership.

A sound deployment begins with requirements, not with a radio purchase. Engineers should identify application deadlines, packet sizes, mobility, availability, safety classification, interference environment and data-governance constraints. The design should then compare private 5G with Wi-Fi, industrial Ethernet, fibre and hybrid architectures. A hybrid design is often preferable: deterministic wired links may serve fixed controllers, while private cellular supports mobile assets and wide-area coverage.

7. Open RAN and Disaggregated Infrastructure

Open RAN seeks greater interoperability and disaggregation among radio, distributed and centralised units, using open interfaces and programmable control. The RAN Intelligent Controller (RIC) concept is commonly associated with near-real-time and non-real-time optimisation functions, including policy, automation and radio-resource applications.

Potential benefits include multi-vendor choice, innovation, automation and deployment flexibility. However, open interfaces do not guarantee plug-and-play interoperability. Performance may be affected by interface overhead, timing requirements, hardware heterogeneity, software maturity and complex responsibility boundaries.

Open RAN also changes the threat model. More interfaces and software components may increase supply-chain risk, misconfiguration risk and opportunities for denial of service. Strong identity, signed software, vulnerability management, independent testing, telemetry and clear vendor accountability are essential. Procurement should evaluate measurable interoperability and performance, not only conformance claims.

8. Semantic and AI-Native Communications

Conventional communication aims to reproduce transmitted bits with low error. Semantic communication instead seeks to convey information relevant to a task or meaning. A semantic encoder may transmit a compact representation of an image, instruction or state, while a decoder reconstructs what is needed for the application.

A general task-oriented objective can be written as

\[

\min_{f,g}\; \mathbb{E}[L(\mathcal{T}(x),g(f(x),z))]+\lambda R(f(x)),

\]

where \(x\) is source data, \(\mathcal{T}(x)\) is the task-relevant target, \(f\) and \(g\) are encoder and decoder, \(L\) is task loss, \(R\) measures communication resource use, and \(\lambda\) weights resource cost. This is a design objective, not a universal physical law. It assumes that the sender and receiver share compatible models and that task loss is measurable.

Semantic systems can reduce transmitted information when exact reconstruction is unnecessary. They can also fail silently if the model hallucinates, becomes stale or interprets context incorrectly. Semantic fidelity must therefore be defined for each application. A reconstructed image that looks plausible may still be unacceptable for medical diagnosis or safety control.

An AI-native network places learning in network operation itself: traffic prediction, beam selection, handover, fault diagnosis, energy optimisation and slice orchestration. A mature architecture must address data quality, distribution shift, explainability, adversarial examples, model provenance, privacy and safe fallback. AI should augment engineering control, not remove accountability.

9. Integrated Sensing and Communication

Integrated sensing and communication (ISAC) uses communication signals, infrastructure or waveforms to infer properties of the environment while providing data connectivity. Sensing outputs may include range, velocity, angle, occupancy or object classification.

For a monostatic radar-like measurement, a delay estimate \(\hat{\tau}\) can be converted to range approximately as

\[

\hat{R}=\frac{c\hat{\tau}}{2},

\]

where \(c\) is propagation speed and the factor two represents a signal path to the object and back. This relation assumes a monostatic reflection geometry and neglects processing bias and multipath. In a one-way communication positioning measurement, the factor would not generally be two.

ISAC creates trade-offs between communication rate, sensing accuracy, transmit power, waveform design, privacy and interference. A base station that detects people or vehicles becomes a sensing system and must address consent, retention, access control and misuse. Regulatory classification may differ depending on whether sensing is passive, active, safety-critical or used for personal identification.

10. Terahertz and Extremely High-Frequency Links

Terahertz research generally concerns frequencies above those traditionally used for millimetre-wave systems, although boundaries vary among communities. The attraction is access to very large contiguous bandwidths. The obstacles include severe path loss, molecular absorption, atmospheric attenuation, blockage, beam misalignment, hardware inefficiency and limited range.

The free-space spreading loss in linear form is represented by the Friis relation

\[

P_r=P_tG_tG_r\left(\frac{\lambda}{4\pi R}\right)^2,

\]

where \(P_t\) and \(P_r\) are transmitted and received powers, \(G_t\) and \(G_r\) are dimensionless antenna gains, \(\lambda\) is wavelength and \(R\) is separation. It assumes far-field propagation, aligned antennas, matched polarisation and free space; atmospheric absorption and blockage require additional loss terms. At a fixed physical antenna aperture, increasing frequency can also change achievable gain, so frequency-only comparisons can be misleading.

Near-term applications are more plausible in short-range indoor links, chip-to-chip or rack-to-rack interconnects, high-resolution sensing and controlled backhaul than in universal wide-area coverage. Link budgets must include realistic antenna efficiency, phase noise, beam-steering loss, atmospheric conditions and installation tolerances.

11. Quantum Networking Research Directions

Quantum networking uses quantum states and entanglement to support capabilities such as quantum key distribution, distributed quantum computation and networked quantum sensing. It is not a replacement for ordinary IP networking and does not imply faster classical communication.

A quantum network requires quantum end nodes, sources, measurement systems, quantum channels and, for long distances, quantum memories and repeaters. Quantum states cannot simply be amplified like classical signals because unknown quantum states cannot be copied perfectly. Entanglement distribution is therefore technologically demanding, with loss, decoherence, synchronisation and memory lifetime as major constraints.

Quantum key distribution can provide security properties based on physical principles under appropriate assumptions, but operational security still depends on authenticated classical channels, device implementation, randomness, endpoint security and correct protocols. Quantum networking research should therefore be evaluated as a layered system, not as a magical replacement for cybersecurity.

A realistic near-term architecture is hybrid: classical networks provide control, authentication and application traffic, while quantum channels provide specialised quantum resources. Interworking, routing, standard interfaces, certification and economic use cases remain open research areas.

12. Illustrative Case: Private 5G with ISAC for a Smart Port

Background. A container port operates autonomous vehicles, cranes, workers’ safety systems and fixed industrial controllers. It requires wide-area mobility, local data processing and detection of objects entering restricted zones.

Problem. Existing Wi-Fi provides insufficient mobility consistency in some yard areas, while a central cloud introduces variable delay. The port also wants to reduce collisions and improve asset tracking without deploying an entirely separate sensing network.

Analysis. Engineers first classify applications. Crane control has stringent availability and bounded latency; video analytics has high throughput but can tolerate some delay; safety alerts require high priority and independent fallback. Coverage surveys identify metallic obstructions, moving containers and interference sources. The sensing requirement is separated into detection, localisation and identification because these have different privacy and accuracy implications.

Approach. The port deploys a private 5G RAN with a local UPF and edge servers. Separate policies are used for control, video and ordinary staff traffic. Packet duplication and redundant transport are considered for selected safety flows. ISAC experiments use existing radio infrastructure to estimate object motion, while certified safety systems retain independent sensors and emergency shutdown logic. Slice-level monitoring reports latency percentiles, loss, availability, handover failures and edge-compute health.

Outcome. The architecture improves mobility management and permits local processing, but it does not eliminate radio blockage or safety risk. The port therefore maintains wired control where feasible, independent emergency mechanisms, strict identity management and human oversight of AI-based detection.

Learning points. Requirements must be application-specific; private 5G is complementary rather than universally superior; sensing data require governance; and safety claims must be demonstrated through system-level verification rather than inferred from a connectivity label.

13. Governance, Security and Professional Practice

Emerging networks combine critical infrastructure, personal data and autonomous decision-making. Key governance concerns include:

• lawful spectrum use and cross-border coordination;
• privacy-preserving localisation and sensing;
• supply-chain transparency and software provenance;
• resilience against outages, jamming and cyberattack;
• fairness in coverage and service allocation;
• energy consumption and lifecycle emissions;
• accessibility and the risk of excluding organisations unable to adopt complex infrastructure;
• auditability of AI decisions and semantic transformations.

Best practice includes threat modelling from the architecture stage, zero-trust identity principles, least privilege, secure boot and signed updates, segmented management planes, independent time sources where appropriate, continuous vulnerability management and tested disaster recovery. Performance claims should identify test conditions, confidence or percentile measures, traffic models and failure assumptions. Researchers should clearly distinguish laboratory demonstrations from deployable system capabilities.

Contemporary Research Agenda

Important research directions include cell-free and distributed MIMO, intelligent metasurfaces, non-terrestrial network integration, energy-aware AI, digital twins for network operations, deterministic wireless, joint communication-computation-sensing design, foundation models for network management, post-quantum security and quantum-classical interoperability. The strongest research questions are interdisciplinary: they combine information theory, wireless propagation, distributed systems, control, machine learning, economics and regulation.

A valuable evaluation framework asks five questions: What is the application objective? What resource is limited? What uncertainty or failure is tolerated? What evidence supports the claimed improvement? What new risks or dependencies are introduced? This prevents technological novelty from being confused with practical value.

Chapter Summary

5G-Advanced extends a cloud-native and programmable foundation toward more intelligent, reliable and context-aware networks. The 5G Core separates control and user planes and exposes network functions through a service-based architecture. Massive MIMO improves spatial reuse but is constrained by channel knowledge, correlation, hardware and energy. URLLC requires analysis of end-to-end latency distributions and reliability, not peak rate alone.

Network slicing provides differentiated logical services over shared infrastructure, while private 5G offers controlled connectivity for industrial environments. Open RAN promotes disaggregation and programmability but introduces interoperability and security challenges. Semantic and AI-native communication optimise task-relevant information and network decisions, yet require safeguards against model error and loss of accountability.

ISAC combines connectivity with environmental inference, creating both technical opportunities and privacy obligations. Terahertz links offer bandwidth but face propagation and hardware limitations. Quantum networking is a specialised research direction requiring hybrid classical control and realistic treatment of loss, authentication and implementation security. Across all paradigms, system-level requirements, evidence-based evaluation, governance and fallback mechanisms are essential.

Key Terms

5G Core: Cloud-oriented core network providing control, user-plane and service functions for 5G systems.

Service-based architecture: Architecture in which network functions expose discoverable services through controlled interfaces.

Massive MIMO: Multi-antenna technology using many antenna elements for spatial multiplexing and beamforming.

URLLC: Communication service class emphasising stringent reliability and latency requirements.

5QI: Identifier associated with defined quality-of-service characteristics in 5G.

Network slicing: Logical partitioning of shared infrastructure into differentiated end-to-end services.

Private 5G: Cellular network operated for a defined organisation, site or user community.

Open RAN: Disaggregated and more openly interfaced radio access network architecture.

RIC: RAN Intelligent Controller supporting policy and optimisation applications.

MEC: Multi-access edge computing placing computation near users or devices.

Semantic communication: Communication focused on task-relevant meaning rather than exact bit reproduction.

AI-native network: Network designed to use machine learning as a fundamental operational capability.

ISAC: Integrated sensing and communication using shared infrastructure or signals for both functions.

Terahertz link: High-frequency wireless link offering potentially very large bandwidth but severe propagation challenges.

Quantum networking: Networking of quantum states or entanglement using quantum channels and specialised nodes.

Packet duplication: Transmission of redundant copies over distinct resources or paths to improve reliability.

Tail latency: High-percentile latency representing slow or extreme delivery cases.

Network function virtualisation: Implementation of network functions using software on virtualised or cloud infrastructure.

Slice assurance: Monitoring and enforcement of service objectives for a network slice.

Zero trust: Security approach requiring continuous verification rather than implicit network-based trust.

Review Questions

Short Answer Questions

1. What is the principal architectural idea of the 5G service-based architecture?

2. Why does massive MIMO performance depend on channel correlation and channel-state information?

3. Distinguish average latency from tail latency in URLLC evaluation.

4. What is the purpose of a local UPF in a private industrial network?

5. State two major propagation challenges for terahertz communication.

Long Answer Questions

1. Explain the major functions of the 5G Core and discuss how control-plane/user-plane separation supports edge computing.

2. Evaluate the benefits and limitations of network slicing for industrial and public-safety services.

3. Discuss Open RAN architecture, its potential advantages and its new security responsibilities.

4. Compare conventional bit-oriented communication with semantic communication, including assumptions and failure modes.

5. Explain how ISAC, terahertz communication and quantum networking represent different research directions rather than a single replacement technology.

Analytical / Application Questions

1. Design a private cellular architecture for a factory containing mobile robots, fixed controllers and high-definition inspection cameras. Justify your choices of core placement, QoS policies, redundancy and edge computing.

2. A proposed URLLC service reports a mean latency of 2 ms but a 99.9th-percentile latency of 80 ms. Analyse why the mean is insufficient and identify measurements and design changes required before deployment.

3. A smart-city project proposes using 5G base stations for continuous pedestrian sensing. Develop a technical and governance assessment covering sensing accuracy, privacy, retention, security, fallback operation and public accountability.

Multiple Choice Questions with Answers

Question 1

Which 5G Core function primarily manages PDU sessions and user-plane selection?

A. AMF

B. SMF

C. AUSF

D. NRF

Correct Answer: B. SMF

Question 2

What is a principal purpose of massive MIMO?

A. Eliminating all propagation loss

B. Providing spatial multiplexing and beamforming

C. Replacing subscriber authentication

D. Guaranteeing zero packet loss

Correct Answer: B. Providing spatial multiplexing and beamforming

Question 3

Why is tail latency important in URLLC?

A. It describes only average throughput

B. Rare excessive delays can violate critical application deadlines

C. It removes the need for reliability measurements

D. It applies only to satellite links

Correct Answer: B. Rare excessive delays can violate critical application deadlines

Question 4

Which component commonly enables local packet breakout in an industrial deployment?

A. Local UPF

B. NRF only

C. AUSF only

D. Remote DNS server exclusively

Correct Answer: A. Local UPF

Question 5

Network slicing is best described as:

A. A mandatory separate physical network for every user

B. Logical differentiation of services over shared infrastructure

C. A radio antenna calibration method

D. A quantum encryption protocol

Correct Answer: B. Logical differentiation of services over shared infrastructure

Question 6

Which is a recognised Open RAN concern?

A. Interfaces automatically guarantee interoperability

B. Disaggregation can increase integration and security complexity

C. Open RAN cannot use software

D. Open RAN eliminates the need for timing synchronisation

Correct Answer: B. Disaggregation can increase integration and security complexity

Question 7

Semantic communication primarily seeks to:

A. Transmit every bit without regard to task

B. Transmit task-relevant information or meaning efficiently

C. Replace all encryption with compression

D. Increase antenna aperture physically

Correct Answer: B. Transmit task-relevant information or meaning efficiently

Question 8

For a monostatic reflection measurement, the ideal range relation is:

A. R=cτ

B. R=2cτ

C. R=cτ/2

D. R=τ/(2c)

Correct Answer: C. R=cτ/2

Question 9

Which is a major challenge for terahertz links?

A. Unlimited range through walls

B. Molecular absorption and blockage

C. Absence of beam alignment requirements

D. Inability to provide bandwidth

Correct Answer: B. Molecular absorption and blockage

Question 10

Why are quantum repeaters researched for quantum networking?

A. To copy unknown quantum states perfectly

B. To address long-distance loss and entanglement distribution limitations

C. To replace all classical authentication

D. To increase classical packet size

Correct Answer: B. To address long-distance loss and entanglement distribution limitations