Reimagining Wireless: A Literature Review of the 6G Cyber-Physical Continuum

Abstract

As the global deployment of fifth-generation (5G) networks matures, the research community is conceptualising sixth-generation (6G) systems, projected for deployment around 2030. This article presents a comprehensive, evidence-based examination of the technological innovations and applications that characterise this transition, informed by a scoping review of 57 sources published between January 2020 and August 2025. The transition to 6G signifies a fundamental transformation from a mere communication utility to an intelligent, sensing, and globally integrated cyber-physical continuum, propelled by a strategic reassessment of the network’s societal function and the practical insights gained from the 5G era. We critically analyse the foundational physical layer technologies that facilitate this vision, including Reconfigurable Intelligent Surfaces (RIS), Terahertz (THz) communications, and the transition to Extremely Large-Scale MIMO (XL-MIMO), emphasising their interdependencies and the fundamental shift towards near-field physics. The analysis encompasses the architectural transformation necessary to address this new complexity, elucidating the principles of the AI-native network, the seamless integration of Non-Terrestrial Networks (NTN) into a cohesive three-dimensional framework, and the functional convergence of communication and sensing (ISAC). We also look at how these changes affect the real world by looking at data from trials and case studies in smart cities, intelligent transportation, and digital health. The article synthesises the overarching challenges in security, sustainability, and scalability, arguing that the path to 6G is defined by two intertwined grand challenges: building a trustworthy and sustainable network. By outlining the critical research imperatives that stem from these challenges, this work offers a holistic framework for understanding how these interconnected developments are evolving wireless networks into the intelligent fabric of a digitised and sustainable society.

1. Introduction

1.1. Rationale

Wireless communication has evolved in a predictable way, with a new generation coming out about every ten years to make big improvements in speed and efficiency. As fifth-generation (5G) networks are rolled out around the world, a wide and active research effort in academia, industry, and government has begun to set the vision and technologies for the sixth generation (6G), which is expected to be available for commercial use around 2030 [1]. The growing agreement is that 6G would not just be an extension of what 5G can do. Instead, it shows a change in the way networks are built, from being mainly for communication to being an intelligent, multi-purpose fabric that connects the physical, digital, and biological worlds.

This big idea has led to a huge amount of writing that is growing quickly. Research is growing quickly in many different technical fields, many of which are still separate from each other. These fields include new physical layer technologies, AI-native architectures, transformative applications, and their effects on society. The extensive quantity and variety of this research necessitate a thorough literature review to integrate these divergent bodies of work, discern principal trends and patterns, critically assess the current State-of-the-Art, and highlight the most urgent unresolved issues and challenges. This paper seeks to furnish a comprehensive and organised overview of the present 6G research landscape [2,3].

The global deployment of fifth-generation (5G) wireless networks has marked a significant milestone, enabling enhanced mobile broadband (eMBB), massive machine-type communications (mMTC), and ultra-reliable low-latency communications (URLLC). These capabilities have begun to power a new wave of applications, from high-definition video streaming to the initial phases of the Internet of Things (IoT) and industrial automation. However, as society and industry accelerate their digital transformation, the architectural and performance limits of 5G are becoming increasingly apparent. The demands of the 2030s envision a world where the digital and physical realms merge into a seamless cyber-physical continuum, requiring a network that is not just faster but fundamentally more intelligent, perceptive, and integrated into the fabric of society.

Future applications such as true holographic communication, city-scale digital twins, the tactile internet, and massive autonomous systems will require performance guarantees that far exceed 5G’s capabilities. These applications demand not only terabit-per-second (Tbps) data rates and sub-millisecond latencies but also extreme reliability, global coverage, and a high degree of contextual awareness. The exponential growth in connected devices, projected to reach trillions, necessitates a network architecture that can manage unprecedented complexity and density while remaining sustainable and energy efficient. Consequently, the evolution towards the sixth generation (6G) is not merely an incremental upgrade but a revolutionary leap, driven by the need to build a platform for a hyper-connected, intelligent, and sustainable world.

1.2. Objectives

This review aims to create a cohesive narrative of the 6G evolution by systematically addressing the following objectives:

• To identify the constraints of 5G that necessitate the transition to 6G.
• To analyse the fundamental technologies and architectural transformations that characterise the 6G vision.
• To evaluate the evidence for its revolutionary applications.
• To synthesise the principal security and sustainability challenges that must be addressed for its successful implementation.

2. Methods

This scoping review was designed and conducted in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) guidelines. A review protocol was established prior to the review, but was not registered.

2.1. Eligibility Criteria

The eligibility criteria required sources to be peer-reviewed journal articles, conference proceedings, or substantive industry and academic white papers published in English. The publication window was set between January 2020 and August 2025 to capture the most recent developments following the widespread deployment of 5G. To be included, the content was required to focus on the vision, enabling technologies, network architecture, applications, or challenges related to 6G wireless systems. Conversely, sources were excluded if they were editorials or non-technical news articles.

2.2. Information Sources and Search Strategy

A comprehensive search was executed in August 2025 across four electronic databases: IEEE Xplore, ACM Digital Library, Scopus, and the ArXiv preprint server. The search strategy was designed to be broad to capture the diverse literature on this topic. The full search string for IEEE Xplore is representative:

((“6G” OR “sixth generation”) AND (“wireless” OR “communication” OR “networks”)) AND (“vision” OR “architecture” OR “terahertz” OR “reconfigurable intelligent surface” OR “RIS” OR “XL-MIMO” OR “non-terrestrial” OR “NTN” OR “integrated sensing and communication” OR “ISAC” OR “AI-native” OR “security” OR “sustainability”)

2.3. Selection of Relevant Articles

All retrieved records were imported into a reference management software, and duplicates were removed. Two reviewers independently screened the titles and abstracts of the remaining records against the eligibility criteria. Any disagreements were resolved through discussion. The full texts of potentially relevant articles were then retrieved and assessed for final inclusion by the same two reviewers.

2.4. Data Charting Process

A data charting form was developed and piloted on a small sample of included articles. One reviewer charted the data from all included sources, and a second reviewer verified the charted data for accuracy and completeness. The following data items were extracted:

• Data Items: Author(s), year of publication, publication type, technology domain (e.g., Physical Layer, Architecture), specific technology/concept (e.g., RIS, AI-native), proposed applications, and identified challenges (e.g., security, energy consumption).

2.5. Synthesis of Results

The charted data were synthesised thematically. An iterative process was used to group related concepts and technologies. The synthesis focused on identifying the key technological pillars, the required architectural evolution, and the overarching challenges, which formed the narrative structure of Section 3.

3. Results

3.1. Selection of Sources of Evidence

The database search initially identified 2512 records. After removing 640 duplicates, 1872 records were screened based on their titles and abstracts, of which 1422 were excluded. The full texts of the remaining 450 articles were assessed for eligibility. Of these, 393 were excluded for reasons such as not being directly relevant to 6G (n = 170), being non-technical in nature (n = 130), or lacking sufficient detail (n = 93). Ultimately, 57 sources met the inclusion criteria and were included in this scoping review. The full PRISMA-ScR flow diagram [4] is presented in Figure 1.

3.2. Characteristics and Synthesis of Sources

• The Evolution of Wireless Communications: From 1G to 6G

To understand the vision for 6G, it is essential to place it within the historical context of wireless evolution. Since the 1980s, a new generation of mobile communication has emerged approximately every decade, each defined by a paradigm-shifting technological advancement and a new set of services that have profoundly reshaped society.

The journey began with the First Generation (1G) in the 1980s, which introduced the concept of mobile voice communication. Using analogue technology, 1G systems were characterised by bulky devices, poor voice quality, and non-existent security, but they represented the revolutionary first step in untethering communication from fixed landlines. The Second Generation (2G), launched in the early 1990s, marked the critical transition from analogue to digital. Based on standards like the Global System for Mobile Communications (GSM), 2G offered improved voice clarity, enhanced security through encryption, and introduced transformative data services such as the Short Message Service (SMS) and later, Multimedia Messaging Service (MMS).

The Third Generation (3G), arriving in the early 2000s, was designed to bring the internet to mobile devices. With technologies like Wideband Code Division Multiple Access (WCDMA), 3G provided significantly higher data rates (up to a few Mbps), making mobile web browsing, video calling, and early application usage a reality. This generation laid the groundwork for the smartphone revolution. The Fourth Generation (4G), introduced around 2008 with its Long-Term Evolution (LTE) standard, supercharged the mobile internet. Based on Orthogonal Frequency Division Multiplexing (OFDM), 4G delivered true broadband speeds (tens to hundreds of Mbps), enabling high-definition video streaming, online gaming, and the rich app-driven ecosystem that defines the modern mobile experience.

The Fifth Generation (5G), with initial deployments starting in 2019, was designed to be more than just a faster network. It expanded the focus beyond human-centric communication to connect a massive number of devices (mMTC) and support critical applications requiring near-instantaneous response times (URLLC). Finally, the vision for Sixth Generation (6G), expected around 2030, is to create a unified platform that integrates communication, sensing, computation, and intelligence. It aims to deliver unprecedented performance while being guided by core societal values, as detailed in the following section. This evolutionary trajectory is summarised in

Table 1. Key Characteristics and Technologies of Wireless Generations (1G–6G).

Generation	Era	Key Technology	Core Service	Peak Data Rate	Latency
1G	1980s	Analogue (FDMA)	Mobile Voice Calls	2.4 Kbps	~Hundreds of ms
2G	1990s	Digital (TDMA, CDMA)	Digital Voice, SMS, MMS	64–200 Kbps	~200–400 ms
3G	2000s	WCDMA, HSPA	Mobile Internet, Video Calling	2–21 Mbps	~100–200 ms
4G	2010s	LTE, OFDM, MIMO	Mobile Broadband, HD Streaming	1 Gbps	~30–70 ms
5G	2020s	NR, Massive MIMO, mmWave	eMBB, URLLC, mMTC	20 Gbps	1 ms
6G	2030s	AI-Native, THz, ISAC, RIS	Immersive XR, Holography, Digital Twins	1 Tbps	0.1 ms

.

Table 1 synthesises the defining features of each wireless generation, illustrating the consistent progression in performance and the evolution of services from basic voice to integrated intelligent systems.

• A Framework for Understanding the 6G Ecosystem

To provide a clear and structured analysis of the multifaceted 6G ecosystem, this paper is organised according to the conceptual framework illustrated in Figure 1. This framework follows a bottom-up approach, starting from the foundational physical layer technologies and building up to the applications and societal values that they enable.

The framework begins with the Foundational Physical Layer, a synergistic triad of technologies—Reconfigurable Intelligent Surfaces (RIS), Terahertz (THz) Communications, and Extra-Large MIMO (XL-MIMO)—that provide the raw performance capabilities of 6G. Building upon this foundation are the Key Enablers, namely the AI-Native Paradigm and Integrated Sensing and Communication (ISAC), which introduce intelligence and perception into the network fabric. These enablers support the diverse and demanding 6G Applications, such as Digital Health, Smart Transportation, and Smart Cities. The entire structure is encapsulated and guided by the overarching principles of Value-Centric Design (KVIs) and the Green Imperative (Sustainability), which dictate the ultimate goals and constraints of the 6G system. This layered approach highlights the interdependencies between different components and reinforces the paper’s central argument that 6G must be understood as a holistic, integrated system.

Figure 1: A Conceptual Framework of the Manuscript. This figure illustrates the logical structure of the paper, showing how foundational physical layer technologies, enabled by AI and ISAC, support advanced applications, all within a guiding framework of societal values and sustainability.

3.2.1. The 6G Imperative: From Connectivity to a Cyber-Physical Fabric

Moving from one wireless generation to the next requires not only major technological advances but also a strategic rethinking of the network’s main purpose. Governments, international standards bodies, and research groups are all working together to shape the vision for 6G as the 5G ecosystem grows. This vision, however, is not a straightforward linear extrapolation of 5G’s performance objectives [5]. Instead, it shows a big change in direction based on what businesses and society learned from the 5G deployment cycle. The path to 6G is marked by a deliberate shift from a technology-centred design, which was driven by the desire to push engineering limits, to a value-centred design, where the needs of society and the economy shape the network’s architecture. This signifies a paradigm shift from designing networks solely for connecting individuals and objects to developing an intelligent, multifunctional framework that seamlessly integrates the digital and physical realms [6].

• Beyond Throughput: The Socio-Economic Drivers for 6G

The pursuit of ambitious, technology-focused Key Performance Indicators (KPIs), like peak data rates of 20 Gbps and air interface latency of 1 ms, was a big reason why 5G was developed. Even though these engineering goals were met, the widespread use of apps that need these extreme capabilities, especially in the Ultra-Reliable and Low-Latency Communications (URLLC) field, has been slower than expected [7]. This experience has shown that there might be a gap between the highest possible performance on isolated metrics and the real value that consumers and businesses receive from them. The 5G era has shown that a network’s success cannot be measured just by its technical specs; it also needs to be able to support services that are good for business and society [8].

In response, a wider range of socio-economic factors are shaping the 6G vision, going beyond just connectivity metrics. A growing number of people agree that the next generation of wireless infrastructure needs to be built to solve important global problems. These include closing the digital divide by making high-speed internet available to everyone, helping to meet the UN’s Sustainable Development Goals (SDGs) for global sustainability [9], and building strong and reliable critical infrastructure for a society that is becoming more digital. As a result, the design imperative for 6G is changing from a “technology-push” model, which is based on what can be achieved with technology, to a “societal-pull” model, which is based on what society needs and what is valuable to the economy.

• The IMT-2030 Vision: A Paradigm Shift in Network Capabilities

The International Telecommunication Union Radiocommunication Sector (ITU-R) is laying the formal groundwork for the 6G vision with its IMT-2030 Framework. This framework, which will guide standards bodies like the 3rd Generation Partnership Project (3GPP), is a big step forward from the 5G paradigm [10]. It builds on the three main use cases for 5G enhanced mobile broadband (eMBB), URLLC, and massive machine-type communications (mMTC), but it puts them into broader, more application-focused groups: immersive communication, hyper-reliable and low-latency communication, and massive communication [9].

More significantly, the IMT-2030 vision introduces new, native network capabilities that signify a qualitative transformation in the network’s role [10]. These are not incremental improvements but fundamental new functions that will be woven into the fabric of the 6G system:

• Ubiquitous Connectivity: This elevates global coverage from a desirable feature to a core usage scenario. It explicitly recognises the need to connect the unconnected and is the primary driver for the native integration of terrestrial and non-terrestrial networks.
• Integrated Sensing and Communication (ISAC): This feature changes the network from a pure communication system into a distributed sensor by combining its functions. The network will use its own radio signals to see, map, and interact with the real world. This will make it possible to offer a new type of cyber-physical service that is aware of its surroundings.
• Integrated AI and Communication: This moves beyond the 5G approach of applying Artificial Intelligence (AI) as an optimisation tool for an existing network architecture. In 6G, AI and Machine Learning (ML) are envisioned as a native component of the network, forming its core operational logic for management, control, and service delivery.

This change in strategy shows that 6G is more than just a faster version of 5G. It is being built from the ground up as a multi-purpose platform that can see its surroundings, learn from them, and provide seamless, intelligent services in space, on land, and in the air. The standardisation roadmap shows this long-term goal. 3GPP Releases 18–20, which are all part of 5G-Advanced, are the first steps towards the first 6G-specific standard, which is expected to be in Release 21 around 2029–2030 [11].

• Redefining Performance: From Key Performance Indicators (KPIs) to Key Value Indicators (KVIs)

The practical lessons of the 5G era have led to a major change in the way network performance is defined and measured. The 6G vision shifts from an exclusive emphasis on isolated, technology-focused KPIs to a more comprehensive and human-centred framework that includes Key Value Indicators (KVIs). KPIs look at how well the network works in terms of technical performance (like data rate in Gbps or latency in ms), while KVIs look at how the network helps with bigger social goals and human values, like security, sustainability, and digital equity. This rebalancing shows that the industry has grown to understand the network’s role as an important part of society’s infrastructure.

This change in thinking is put into action by raising a number of “soft” requirements, which were once seen as secondary design goals in 5G, to the level of primary, measurable goals for 6G. This has big effects on the network’s architecture because the design choices are no longer just about technical factors like latency and throughput. Now, they require complicated, multi-dimensional optimisations between performance, energy use, security, and openness.

• Sustainability: Energy performance is now a top design constraint, not just a qualitative goal. Industry and research organisations are pushing hard for the creation of quantitative and measurable standards for energy efficiency. The goal is to create a system that is not only more efficient per bit but also helps to lower the ICT sector’s overall energy footprint. Both environmental concerns and the high operational costs (OPEX) of powering networks that are becoming more complicated are driving this. The objective is to develop a “Green 6G” system that incorporates energy awareness into its fundamental operational logic [12,13,14].
• Trustworthiness, Security, and Resilience: Security is being treated as a basic design principle, not an afterthought. This is because the network is so closely linked to important infrastructure, personal data, and autonomous systems. Ten countries, including the US, UK, and Japan, made a joint statement in support of a set of principles for 6G. These principles say that 6G must be “Secure, Open, and Resilient by Design.” This makes trustworthiness, data privacy, and resilience core architectural requirements, which means that we need to move to new security models like Zero-Trust Architecture [15,16].
• Digital Inclusion and Equity: The IMT-2030 framework’s inclusion of “ubiquitous connectivity” as a primary usage scenario makes bridging the digital divide a core objective of the 6G mission. 6G is expected to be the first wireless generation that was designed from the ground up to give everyone, no matter where they live or how much money they have, affordable, accessible, and fair connectivity. This is the main reason why terrestrial networks and non-terrestrial platforms work together so well. It is a big change from 5G [10].

The shift from a technology-focused design to one that focuses on value is probably the most important thing that will happen with 6G. People will judge 6G’s success not only by how well it works technically, but also by how well it can provide services that are fair, reliable, and long-lasting. This requires architectures that are not only powerful but also inherently adaptable, efficient, and secure by design.

Table 2. Evolution of Key Performance and Value Indicators from 5G to 6G.

Parameter	5G (IMT-2020)	6G (IMT-2030 Vision)
Core Design Philosophy	Connecting People and Things	Creating a cyber-physical continuum, a multi-purpose sensing and communication fabric
Primary Metrics	Technology-centric Key Performance Indicators (KPIs): Peak Data Rate (20 Gbps), Latency (1 ms), Connection Density (106 devices/km2)	Value-centric Key Value Indicators (KVIs) + enhanced KPIs: Sustainability, Trustworthiness, Security, Digital Inclusion
Key Drivers	Technology-push: Maximising engineering capabilities (e.g., eMBB, URLLC, mMTC)	Value/Societal-pull: Addressing societal needs and enabling new commercial value (e.g., Immersive Communication, Ubiquitous Connectivity)
Security Approach	Primarily an overlay; security features added to the architecture	Secure-by-Design: Trustworthiness as a foundational, non-negotiable architectural principle
Sustainability Approach	A secondary design goal, focused on improving energy efficiency per bit	A primary design constraint (“Green 6G”), focused on both efficiency per bit and reducing total network energy consumption
Coverage Goal	Best-effort terrestrial coverage, with NTN as a separate, non-integrated system	Ubiquitous global coverage as a core usage scenario, achieved through the native integration of terrestrial and non-terrestrial networks
AI/ML Integration	Applied as an optimisation tool for existing network functions	AI-native: AI/ML forms the core operational logic of the network for orchestration, management, and control

shows how the main philosophical and technical changes from the 5G to the 6G vision are different.

3.2.2. Foundational Physical Layer Technologies

The ambitious vision for 6G is predicated on a suite of revolutionary physical layer technologies that fundamentally alter how wireless networks interact with the electromagnetic environment. These innovations are not merely incremental improvements; they represent paradigm shifts that enable unprecedented data rates, intelligent environmental control, and new spatial processing capabilities. However, each of these technologies introduces its own set of formidable physical and engineering challenges. Their viability depends not only on individual breakthroughs but also on their synergistic integration into a coherent system.

• The Co-Dependent Triad: RIS, THz Communications, and XL-MIMO

No single technology can meet all the demands of 6G. Instead, its physical layer is likely to be built upon the tight integration of several key innovations that mutually compensate for each other’s weaknesses. This co-dependent relationship is particularly evident in the triad of THz, XL-MIMO, and RIS.

Terahertz (THz) Communications: Operating in the frequency range of 0.1–10 THz, this technology unlocks vast, contiguous blocks of spectrum, offering the ultra-high bandwidth necessary to achieve multi-gigabit and even terabit-per-second data rates. This raw capacity is a fundamental enabler for immersive applications like holographic telepresence and real-time digital twins. However, the primary drawback of THz waves is their extremely poor propagation characteristics. They suffer from severe free-space path loss and are highly susceptible to atmospheric absorption and blockage by common obstacles, making reliable communication over distance a significant challenge.

Extra-Large MIMO (XL-MIMO): As a natural evolution of massive MIMO, XL-MIMO involves deploying antenna arrays with hundreds or even thousands of elements. These massive apertures can generate highly focused, pencil-like beams, providing enormous beamforming gains that can compensate for the severe path loss inherent in THz communications. By concentrating energy directly toward the user, XL-MIMO makes THz links feasible over practical distances. However, XL-MIMO’s reliance on narrow beams makes it extremely vulnerable to line-of-sight (LoS) blockage. If the direct path is obstructed, the communication link can be completely severed.

Reconfigurable Intelligent Surfaces (RIS): RIS technology addresses the critical vulnerability of LoS blockage shared by both THz and XL-MIMO. An RIS is a metasurface composed of a large number of low-cost, passive elements that can be electronically controlled to manipulate incident electromagnetic waves. By precisely adjusting the phase shift of each element, an RIS can reflect a signal in a desired direction, effectively creating a virtual LoS path around an obstacle. This capability is crucial for maintaining connectivity in complex environments. Because they are nearly passive, RISs are also highly energy-efficient and can be deployed flexibly on surfaces like walls, ceilings, and billboards.

The synergy is clear: THz provides the bandwidth, XL-MIMO provides the beamforming gain to make the link viable, and RIS provides the reliability by steering the signal around obstacles. Each technology mitigates a fundamental weakness of the others, forming a robust and powerful foundation for the 6G physical layer.

• Reconfigurable Intelligent Surfaces (RIS): Engineering a Smart Radio Environment

For decades, the wireless propagation channel has been treated as an uncontrollable, stochastic element whose detrimental effects, such as fading and blockage, must be compensated for at the transmitter and receiver. Reconfigurable Intelligent Surfaces (RIS), also known as Large Intelligent Surfaces (LIS), represent a radical departure from this paradigm, transforming the channel from a passive medium into an active, controllable component of the network itself.

An RIS is a two-dimensional meta surface composed of a large number of low-cost, nearly passive reflecting elements, often called unit cells. These elements are engineered from composite materials and can be electronically controlled to manipulate the properties of incident electromagnetic (EM) waves. By embedding simple switching components like PIN or varactor diodes into each unit cell, the impedance of the cell can be altered, which in turn changes the phase shift and/or amplitude of the reflected wave. By orchestrating the phase shifts across the thousands of elements on the surface in real time, an RIS can effectively steer the reflected signal’s energy in a desired direction, a process known as passive beamforming.

This capability to create a “smart radio environment” yields two primary benefits. First, it enables coverage enhancement, particularly in high-frequency bands like millimetre-wave (mmWave) and THz, where signals are highly susceptible to blockage. An RIS can be deployed on a building facade to intercept a signal and reflect it around an obstacle to a user, effectively establishing a virtual line-of-sight (LoS) path where none existed before. Second, by intelligently focusing the reflected signal’s energy towards a specific user, an RIS can significantly improve the received signal-to-noise ratio (SNR), leading to higher data rates and improved reliability with very low power consumption compared to traditional active relays [17].

To formalise this, consider a basic RIS-assisted multiple-input single-output (MISO) communication system. The received signal, y, at the user can be modelled as the superposition of the signal from the direct link (base station to user) and the signal from the RIS-reflected link (base station to RIS to user). The mathematical representation is given by:

$$\mathrm{y}=({\mathrm{h}}_{\mathrm{d}}^{\mathrm{H}}+{\mathrm{h}}_{\mathrm{r}}^{\mathrm{H}}\mathsf{\Phi }\mathrm{G})\mathsf{{\rm X}}+\mathrm{n}$$

Here, $h_d\in C^{M\times 1}$ is the channel vector for the direct link, G∈CN×M is the channel matrix from the base station (with M antennas) to the RIS (with N elements), $h_r\in C^{N\times 1}$ is the channel vector from the RIS to the user, $\mathsf{{\rm X}}$ is the transmitted signal, and n is the additive white Gaussian noise. The key controllable component is the RIS phase-shift matrix, $\mathsf{\Phi }=diag(e^{j{\theta }_1},e^{j{\theta }_2},\dots ,e^{{j\theta }_N})$, where each ${\theta }_n$ is the phase shift applied by the $n$-th element. The primary optimisation problem is to maximise the received SNR by jointly designing the active beamforming vector at the base station and the passive phase shifts in $\mathsf{\Phi }$, a problem that is non-convex and computationally challenging to solve in real time [18]. The practical deployment of RIS technology hinges on solving these formidable challenges. Chief among them is the development of robust and low-overhead channel estimation strategies to acquire the necessary channel state information (CSI) for the cascaded BS-RIS-user link, and the design of scalable, real-time control mechanisms to update the phase shifts for mobile users [19].

• Reconfigurable Intelligent Surfaces (RIS) vs. Active Relays: A Comparative Study

Active relays, which amplify-and-forward (AF) or decode-and-forward (DF) signals, have been used in cellular networks to extend coverage and improve performance for cell-edge users. RISs serve a similar purpose but operate on a fundamentally different principle: they reflect and refocus signals passively rather than actively regenerating and transmitting them. This distinction leads to significant trade-offs in performance, cost, and efficiency, as detailed in

Table 3. Comparative Analysis of RIS vs. Active Relays.

Metric	Active Relay (Amplify/Decode-and-Forward)	Reconfigurable Intelligent Surface (RIS)
Signal Processing	Actively processes and regenerates the signal (amplification, decoding, re-encoding).	Passively reflects and phase-shifts the incident signal. No active processing.
Power Consumption	High. Requires active RF chains, power amplifiers, and a continuous power supply.	Very Low. Consumes minimal power, mainly for the control circuitry of the elements.
Hardware Cost	High. Involves complex and expensive components like ADCs/DACs and power amplifiers.	Low. Composed of simple, low-cost passive reflecting elements and a controller.
Spectral Efficiency	Can be very high, especially with advanced processing (e.g., DF).	Dependent on the number of elements and path loss; can be limited by the double path loss effect.
Noise	Introduces additional thermal noise (especially AF relays).	Does not introduce thermal noise, leading to a cleaner signal reflection.
Self-Interference	A major challenge for full-duplex relays, requiring complex cancellation techniques.	Not applicable, as RIS is a passive reflector and does not transmit its own signal.
Deployment Flexibility	Limited by the need for a dedicated power supply and often fibre backhaul.	High. Can be deployed on various surfaces (walls, ceilings) with minimal power requirements.

.

The primary advantage of RIS lies in its energy efficiency and low hardware cost. Because RISs do not require active radio frequency (RF) chains, power amplifiers, or complex digital signal processing, their power consumption is orders of magnitude lower than that of active relays, which require a constant power supply for their electronic components. This makes RIS a highly scalable and sustainable solution, especially for dense deployments. Furthermore, the absence of active components means RISs do not introduce additional thermal noise into the system, unlike AF relays, which amplify both the signal and the noise.

However, active relays possess capabilities that passive RISs lack. They can perform full signal processing, including decoding, re-encoding, and interference cancellation, which can lead to higher spectral efficiency in certain scenarios. Full-duplex active relays can transmit and receive simultaneously, though they must contend with self-interference, a problem that passive RISs do not face. The key performance limitation of RIS is the “double path loss” effect: the signal power attenuates over both the base-station-to-RIS link and the RIS-to-user link.

For an RIS to be effective, it must be deployed strategically to minimise the total path loss, and the number of reflecting elements must be large enough to provide sufficient passive beamforming gain.

The research gap lies in developing low-complexity, scalable control and optimisation algorithms for RISs with thousands of elements, as well as hybrid RIS/relay systems that combine the energy efficiency of passive reflection with the signal processing power of active amplification.

Table 3 highlights the fundamental trade-offs between active relays and RISs, showing that RIS offers a paradigm of low-cost, energy-efficient coverage extension at the expense of signal processing capabilities.

• Terahertz (THz) Communications: Unlocking Unprecedented Bandwidth Amidst Physical Constraints

The Terahertz (THz) frequency band, spanning from 0.1 to 10 THz, is widely regarded as a cornerstone technology for 6G, primarily for one reason: the availability of vast, contiguous blocks of untapped bandwidth [20]. This enormous spectral resource is essential for enabling futuristic applications that demand data rates reaching hundreds of gigabits-per-second (Gbps) or even terabits-per-second (Tbps), such as holographic communication, real-time digital twins, and high-resolution environmental sensing. However, the transition to these ultra-high frequencies is fraught with fundamental physical and technological challenges that currently limit its practical application.

The primary obstacle is severe propagation loss, which has two dominant components [21]. The first is free-space path loss (FSPL), which dictates the signal attenuation due to the geometric spreading of the wavefront. According to the Friis transmission formula, FSPL increases with the square of both the distance (d) and the frequency ($f$). This $f^2$ dependency means that, all else being equal, a signal at 300 GHz will experience 100 times (or 20 dB) more path loss than a signal at 30 GHz over the same distance [22]. The second, and more critical, component is molecular absorption loss. THz waves have frequencies that coincide with the rotational resonance frequencies of molecules in the atmosphere, particularly water vapour ($H_{2}O$) and oxygen ($O_2$). This causes the signal’s energy to be absorbed by these molecules, leading to extremely high attenuation in specific frequency bands [22]. This phenomenon creates “transmission windows” where communication is viable but limits the effective communication distance to metres or tens of metres, especially in outdoor or humid environments.

Compounding the propagation challenges is the relative immaturity of THz hardware, often referred to as the “THz gap” [20]. Compared to mature microwave and mmWave technologies, the development of cost-effective and powerful THz-band components is still in its early stages [23]. Current THz transceivers have limited output power, and the required semiconductor materials (e.g., III-V compounds like Gallium Arsenide) and photonic sources are expensive and complex to manufacture, posing significant barriers to mass-market deployment [20]. To compensate for the high path loss, THz systems must employ highly directional antennas to focus energy into narrow beams. While this improves the link budget, it makes the communication link extremely sensitive to alignment and highly vulnerable to blockage by any physical obstacle, including the human body [24].

• The Evolution of Multi-Antenna Systems: From Massive MIMO to Extremely Large-Scale Apertures

Massive Multiple-Input Multiple-Output (MIMO), a cornerstone of 5G that uses large antenna arrays at the base station to serve multiple users simultaneously, is evolving along several trajectories for 6G, pushing the boundaries of spatial processing and fundamentally altering the physics of propagation.

One emerging paradigm is Cell-Free (CF) Massive MIMO. This approach challenges the traditional cellular concept by eliminating cell boundaries. Instead of dividing an area into cells, each served by one base station, CF M-MIMO deploys a large number of distributed, single-antenna access points (APs) that are connected to a central processing unit. These APs cooperate to coherently serve all users in their coverage area, mitigating the classic problem of poor performance for users at the cell edge and providing a more uniform quality of experience [25].

A second, more profound evolution is the quantitative leap in scale to Extremely Large-Scale MIMO (XL-MIMO), also referred to as Ultra-Massive MIMO (UM-MIMO). This involves increasing the number of antenna elements from the hundreds typical in 5G to thousands or even tens of thousands in 6G [26]. This massive increase in the physical aperture of the antenna array fundamentally changes the nature of the electromagnetic field. Communication transitions from the far-field, where the distance between the transmitter and receiver is large enough for the wavefront to be approximated as a planar wave, to the near-field, where the curvature of the wavefront becomes significant and cannot be ignored [27,28].

The boundary between these two regions is commonly defined by the Rayleigh distance, given by $Z_R={\displaystyle \frac{{2D}^2}{\lambda }}$, where $D$ is the largest dimension of the antenna array and $\lambda$ is the wavelength. As the antenna aperture $D$ increases (with XL-MIMO) [29] and the wavelength $\lambda$ decreases (at higher frequencies), the Rayleigh distance can extend from a few metres to hundreds of metres, meaning that many users will be located within the near-field of the base station [28]. This transition necessitates a shift from the far-field planar wave model to the more accurate near-field spherical wave model. In the far-field model, the channel response depends only on the angle of arrival/departure. In the near-field spherical model, the channel response depends on both the angle and the distance between the transmitter and receiver. This additional degree of freedom in the distance dimension enables a new capability beyond the beam-steering of far-field systems: beam-focusing, which allows the array to concentrate signal energy at a specific point in 3D space, offering unprecedented spatial resolution and multiplexing capabilities [28]. The conceptual endpoint of this evolution is Holographic MIMO [30], which envisions an antenna array so dense that it can be modelled as a continuous electromagnetic surface, allowing for near-complete control over the shape of the radiated field.

The foundational technologies of the 6G physical layer are not independent innovations but rather a tightly co-dependent triad. The extreme path loss inherent to THz communication renders it impractical without a mechanism to achieve massive beamforming gain. This is precisely the capability provided by XL-MIMO, whose thousands of antenna elements can generate the highly focused energy beams required to close the THz link budget. Thus, XL-MIMO is a direct enabler for THz communication [31]. However, both of these technologies share a critical vulnerability: their reliance on highly directional, pencil-like beams makes them extremely susceptible to line-of-sight blockage. This is where RIS becomes essential. By intelligently redirecting signals, RIS can create virtual line-of-sight paths, effectively navigating around obstacles that would otherwise sever the link. This symbiotic relationship means that a practical 6G physical layer cannot be realised by considering these technologies in isolation; they must be designed and analysed as an integrated system where each component mitigates the fundamental weaknesses of the others.

Table 4. Analysis of 6G Enabling Physical Layer Technologies.

Technology	Core Principle	Key Advantage for 6G	Primary Open Research Challenge
Reconfigurable Intelligent Surfaces (RIS)	Transforming the wireless channel into a controllable, software-defined environment through passive, phase-shifting meta-surfaces.	Low-power, low-cost coverage extension and performance enhancement by creating virtual line-of-sight paths and passive beamforming.	Cascaded Channel Estimation: Developing low-overhead techniques to accurately estimate the channel state information for the combined BS-RIS-user link in real time for mobile environments.
Terahertz (THz) Communications	Utilising the vast, contiguous bandwidth available in the 0.1–10 THz frequency range.	Enables unprecedented data rates (approaching Tbps), essential for futuristic applications like holographic communication and high-resolution sensing.	Overcoming Propagation Loss: Mitigating severe free-space path loss and molecular absorption, coupled with the immaturity and high cost of THz hardware components (“THz Gap”).
Extremely Large-Scale MIMO (XL-MIMO)	Scaling antenna arrays to thousands of elements, transitioning propagation physics from the far-field to the near-field.	Provides massive beamforming gain to overcome path loss (enabling THz) and enables beam-focusing in the near-field for high-resolution spatial multiplexing.	Near-Field Channel Modelling and Complexity: Developing accurate yet computationally tractable channel models for the spherical wave region and designing efficient signal processing algorithms for arrays with thousands of elements.

provides an analysis of these key enabling technologies.

• Extra-Large MIMO (XL-MIMO) vs. Cell-Free Massive MIMO

XL-MIMO and Cell-Free Massive MIMO both leverage a very large number of antennas to serve users, but they differ fundamentally in their deployment topology. XL-MIMO typically refers to a co-located or centralised deployment where an extremely large antenna array is installed at a single site, such as a base station. In contrast, Cell-Free Massive MIMO distributes a large number of access points (APs), each with one or a few antennas, over a wide geographical area.

These APs are connected via a high-capacity fronthaul network to a central processing unit (CPU) and cooperate to serve all users in the area jointly. This architectural difference leads to the trade-offs summarised in

Table 5. Comparative Analysis of XL-MIMO vs. Cell-Free Massive MIMO.

Metric	Extra-Large MIMO (XL-MIMO)	Cell-Free Massive MIMO
Architecture	Centralised. A very large, co-located antenna array at a single base station.	Distributed. A large number of geographically distributed access points (APs) connected to a central CPU.
Key Benefit	Unprecedented beamforming gain and spatial resolution (beam focusing). High peak capacity.	Superior macro-diversity and uniform coverage. Eliminates cell-edge effects and improves user fairness.
Near-Field Effects	Prominent. Spherical wavefronts allow for focusing on both angle and distance.	Generally, not a primary factor, as individual APs have few antennas.
Backhaul Requirement	Standard backhaul from the base station to the core network.	Extremely high-capacity, low-latency fronthaul from all APs to the central processing unit.
Interference Pattern	Inter-cell interference from neighbouring XL-MIMO base stations.	No cell boundaries: interference is managed through joint processing across all APs.

.

XL-MIMO’s primary advantage is its ability to generate extremely high-resolution beams, a phenomenon known as “beam focusing.” As the array size becomes very large, the electromagnetic field transitions from the far-field (planar wavefronts) to the near-field (spherical wavefronts), allowing for focusing not just in angle but also in distance. This provides unprecedented spatial multiplexing capabilities and can significantly boost the capacity for users located close to the array. However, like any centralised system, it is still susceptible to LoS blockage and can create coverage holes for users far from the base station.

Cell-Free Massive MIMO, on the other hand, excels at providing uniform coverage and eliminating the concept of cell edges. Since users are typically close to several APs, the system can exploit macro-diversity to combat shadow fading and provide a consistently high quality of service regardless of the user’s location. This leads to a significant improvement in fairness and cell-edge user performance compared to traditional cellular systems. The main challenge for Cell-Free systems is the immense fronthaul requirement, as the signals from all distributed APs must be sent to the CPU for coherent joint processing. Managing interference and designing scalable signal processing algorithms are also critical research areas.

A promising future direction involves the convergence of these two paradigms: a Cell-Free architecture where the distributed APs are themselves equipped with XL-MIMO arrays. Such a system could potentially combine the uniform coverage and macro-diversity benefits of Cell-Free with the high spatial resolution and capacity of XL-MIMO, though it would exacerbate the challenges of fronthaul and computational complexity.

Table 5 contrasts the centralised and distributed approaches to scaling MIMO, highlighting their distinct advantages in terms of peak capacity versus uniform coverage.

3.2.3. Architectural Metamorphosis: Moving Towards a Smart, Everywhere, and Converged Network

The revolutionary features added at the physical layer require a complete change in the network’s structure. The International Telecommunication Union Radiocommunication Sector (ITU-R) recently adopted the International Mobile Telecommunications (IMT)-2030 framework. This framework envisions 6G networks that will provide intelligent, seamless connectivity that supports reliable, sustainable, and resilient communications. A 6G system is too big, too complicated, and too dynamic to be managed by hand. It is made up of many different types of networks, both on the ground and in space, and it operates over a wide range of frequencies, from sub-6 GHz to THz. It also has new controllable elements like RIS. This pushes for a change in architecture towards a system that is smart by design, has a three-dimensional reach, and has a single purpose that combines all of its functions. Every new physical layer technology is like a “complexity ratchet,” locking in a new level of operational challenge that needs a more advanced architectural solution to handle it [32].

• The AI-Native Architecture: A Self-Orchestrating Network Nervous System

The need for an “AI-native” architecture comes directly from how hard it is to manage the 6G physical layer’s complexity. Real-time optimisation of thousands of RIS phase shifts, management of fragile THz links, and near-field beam-focusing for mobile users are all tasks that are much too hard for people or normal algorithms to do. So, AI and ML need to be built into the network itself, changing from a tool for improving performance, like in 5G, to the main logic for how 6G works. Researchers often conceptualise this AI-native architecture as a four-layer, closed-loop system that works like the nervous system of the network [33]:

• Intelligent Sensing Layer: This basic layer collects a lot of different types of data from every network element (base stations, UEs, IoT devices, RIS controllers) all the time. This includes radio-frequency data (like CSI and spectrum occupancy), network performance metrics (like latency and throughput), and user context (like location and mobility patterns).
• Data Mining and Analytics Layer: This layer is like the brain of the network; it takes in all the data from the sensing layer. It uses advanced AI and machine learning to find patterns, make predictions, and figure out how the network and its users work, turning raw data into useful information.
• Intelligent Control Layer: This is the network’s decision-making and action hub. It uses information from the analytics layer to make decisions and take action. It uses AI methods, especially reinforcement learning, to keep improving and reach a high level of automation, such as self-configuration, self-optimisation, and self-healing.
• Smart Application Layer: This is the layer that delivers services. It uses the intelligence of the lower layers to offer personalised, context-aware services. It also creates an important feedback loop by checking how well services are working and sending that information back to the control and analytics layers so they can keep becoming better.
• This intelligence cannot all be in the cloud because many 6G apps need very low latency. It has to be spread out over the network. You can train big AI models in the cloud for tasks that do not need to happen in real time, like long-term network planning. For real-time inference and decision-making, like predictive mobility management or dynamic resource allocation, you can use lightweight models on Mobile Edge Computing (MEC) servers at the network edge [34].

Federated Learning (FL) is a key part of this distributed intelligence. Federated Learning (FL) is a machine learning method that protects privacy by training a shared global model across many distributed nodes, like UEs or edge servers, without having to centralise the raw training data [35]. In the FL lifecycle, the central server sends the nodes the first model. After that, each node trains the model on its own private data. Instead of sending the data back, the nodes only send the central server their updated model parameters (like gradients). The central server then combines these to make a better global model. This process is performed over and over again. This method lets the network use the data of all its users while keeping private information private. This is a key part of trustworthy AI in 6G.

• Defining the Future: AI-Native vs. AI-Integrated Architectures

To appreciate the transformative potential of AI in 6G, it is crucial to distinguish between AI-integrated and AI-native architectures.

AI-Integrated (or “Bolted-On”) Architecture: This describes the approach largely taken in 5G. In this model, AI/ML algorithms are added as an overlay to a traditional, human-designed network architecture. AI is used to optimise specific functions or automate existing processes—for example, predicting traffic loads to adjust resources or detecting anomalies in network performance. The underlying network protocols, interfaces, and overall structure remain largely unchanged and are not inherently designed to leverage AI. The intelligence is an “afterthought,” retrofitted onto a legacy framework.

AI-Native Architecture: This is the visionary approach for 6G. In an AI-native system, AI is not an add-on but a fundamental pillar of the network’s design from the ground up. The entire network, from the physical layer to the service orchestration layer, is designed for and with AI. This means that AI models are not just optimising parameters but are integral to the core functionalities. For example, the 6G air interface may be designed by an AI model rather than being based on fixed mathematical constructs. The network is envisioned as a self-evolving system capable of autonomous service creation, zero-touch management, and intent-based orchestration, where human operators specify what they want the network to achieve (the intent), and the AI-native fabric determines how to achieve it in real time.

This transition from AI-integrated to AI-native is not merely a technical upgrade; it represents a fundamental shift in the operational philosophy of telecommunications. It moves the industry from managing static, complex networks through manual configuration to orchestrating dynamic, intelligent, and autonomous systems. This has profound implications, requiring new skill sets focused on data science and MLOps, new infrastructure for distributed training and inference at the network edge, and new frameworks for ensuring the ethical governance and reliability of these autonomous systems.

• Fuelling the AI Engine: The Role of Advanced 6G Datasets

An AI-native network is fundamentally data-driven. The performance, reliability, and intelligence of its AI models are directly determined by the quality and scale of the datasets used for their training and validation. Recognising this, the research community has begun to develop large-scale, multi-modal datasets specifically designed for 6G research, which are essential for moving beyond purely theoretical models to solutions grounded in real-world complexities. Two prominent examples are DeepSense 6G and SynthSoM.

DeepSense 6G: This is a pioneering large-scale, real-world dataset that captures co-existing multi-modal sensing and communication data [36]. It comprises over a million data samples collected from more than 30 diverse scenarios, including vehicle-to-infrastructure (V2I), drone communication, and indoor environments. The key feature of DeepSense 6G is its multi-modality; it provides synchronised data from mmWave communication links, high-resolution cameras, LiDAR, radar, and GPS. This richness enables researchers to develop and validate advanced AI models for applications at the intersection of communication, sensing, and positioning, such as using visual or LiDAR data to predict beam blockages and improve communication reliability.

SynthSoM (Synthetic Synesthesia of Machines): While real-world datasets like DeepSense 6G are invaluable for capturing environmental complexity, they can be expensive to collect and may not cover all possible scenarios. The SynthSoM dataset addresses this gap by providing a high-fidelity, synthetic multi-modal sensing and communication dataset [37]. Generated using a sophisticated simulation platform that integrates multiple high-precision software tools, SynthSoM covers diverse air-ground cooperative scenarios under various conditions (weather, time of day, agent density). The scale of the dataset is comprehensive, encompassing multiple data modalities, including 140 K sets of channel matrices, 18 K sets of path loss, 136 K sets of mmWave radar waveforms (with 38 K radar point clouds), 145 K RGB images, 290 K depth maps, and 79 K sets of LiDAR point clouds. It includes RF channel data, mmWave radar data, and non-RF sensory data like RGB images and LiDAR point clouds. Synthetic datasets offer the advantage of providing perfectly labelled, comprehensive data that can be used to train AI models for scenarios that are rare or dangerous to replicate in the real world, thus complementing and accelerating research.

• AI-Driven Network Management for a Green 6G

One of the most critical applications of the AI-native paradigm is in addressing the KVI of sustainability. As network traffic and complexity grow, AI-driven management is essential for optimising energy consumption and creating a “Green 6G”. AI algorithms can enhance energy efficiency across all network layers by enabling intelligent and proactive resource management. This includes predicting traffic patterns to dynamically activate or deactivate network components, optimising power control in real time, and managing resource allocation to minimise energy overhead while meeting quality of service (QoS) requirements.

The convergence of multiple 6G technologies, orchestrated by AI, offers powerful new avenues for energy optimisation. A compelling State-of-the-Art example is the use of AI to manage complex systems that integrate RIS, ISAC, and advanced multiple access schemes. For instance, recent research has explored using Deep Reinforcement Learning (DRL) to maximise the energy efficiency of an RSMA-IRS-assisted ISAC system. In this scenario, a DRL agent, such as one based on the Proximal Policy Optimisation (PPO) algorithm, learns to jointly optimise the transceiver beamforming at the base station and the phase shifts at the IRS. The agent’s goal is to maximise the overall energy efficiency (bits per Joule) while satisfying the QoS constraints for communication users and the sensing requirements for target detection. This work provides a concrete demonstration of how AI can navigate the complex trade-offs in a multi-technology 6G environment to achieve sustainability goals, moving the concept of Green 6G from a high-level vision to a practical implementation strategy.

• The Next Frontier: Foundation Models and LLMs for the Physical Layer

The recent, dramatic success of Large Language Models (LLMs) like GPT in natural language processing has inspired a new direction of research in wireless communications: the development of foundation models [38]. A foundation model is a large AI model trained on a massive amount of unlabelled data that can be adapted to a wide range of downstream tasks with minimal fine-tuning. This approach contrasts with traditional AI, which typically involves training smaller, specialised models for each task.

To further highlight the advantages of this approach, it is crucial to contrast it with conventional deep learning (DL) models. Conventional DL, while highly effective for specific applications, typically requires designing and training bespoke models for each task. This task-specific paradigm has significant limitations for a system as complex as 6G. Firstly, it is resource-intensive, requiring extensive, time-consuming, and costly retraining cycles whenever new services, topologies, or environmental dynamics are introduced. Secondly, these models are often data-inefficient for their specific task and struggle to generalise to unseen scenarios, lacking the broad reasoning capabilities of larger models. LLMs and foundation models, in contrast, offer a paradigm shift toward operational flexibility and autonomy [38]. Their key advantages lie in (1) strong generalisation, enabling “zero-shot” or “few-shot” learning to handle tasks they were not explicitly trained for; (2) resource-efficient adaptability, where pre-trained knowledge is leveraged, and models are adapted via lightweight fine-tuning or “prompt engineering” rather than full retraining; and (3) multi-modal reasoning, allowing them to process and integrate diverse data streams (e.g., network KPIs, sensor data, user requests) to perform high-level resource allocation and network management [38,39].

The vision is to create a Wireless Physical-layer Foundation Model (WPFM) that, once pre-trained on vast datasets of wireless signals and channel measurements, could be fine-tuned for diverse tasks such as channel estimation, signal detection, waveform design, and beam prediction. Such a model could learn the fundamental “language” of wireless propagation and communication, enabling it to generalise across different environments and scenarios far more effectively than task-specific models.

Furthermore, LLMs are being explored for higher-level network orchestration and management [38]. A 6G-oriented LLM agent, deployed at a base station, could understand high-level service requirements described in natural language, perceive the complex radio environment through multi-modal inputs, and autonomously orchestrate the network’s resources and physical layer functions to fulfil the task optimally [38,39]. This involves training general-purpose LLMs with domain-specific knowledge from sources like 3GPP specifications and research papers to create a “field basic model” that can then be fine-tuned for specific roles. While significant challenges remain, including the design of effective pre-training tasks for wireless data, managing the computational cost of these massive models, and ensuring their reliability, the potential for foundation models and LLMs to bring a new level of adaptability, generalisation, and autonomy to 6G networks is immense.

• The Third Dimension: Native Integration of Non-Terrestrial Networks (NTN)

A major change in 6G architecture is the native integration of Non-Terrestrial Networks (NTN), which makes a single, three-dimensional network fabric. This encompasses Low Earth Orbit (LEO), Medium Earth Orbit (MEO), and Geostationary (GEO) satellites, in addition to High-Altitude Platform Stations (HAPS) and Unmanned Aerial Vehicles (UAVs) [40]. 5G treats NTN as a separate, non-integrated system, but 6G sees it as a “network of networks” where both terrestrial and non-terrestrial components work together as one. This integration is the key to making the 6G vision of truly global, always-on connectivity a reality. It will enable use cases like closing the digital divide in remote areas, making sure that services stay up during natural disasters, and supporting vertical industries like maritime shipping and precision agriculture. But making this unified 3D network is a big technical challenge that needs to be solved at the architectural level [41].

• Large Doppler Shifts: LEO satellites move so fast (up to 7.8 km/s) that they can cause Doppler shifts of tens of kilohertz, which are much larger than those seen in networks on the ground. This can make it very hard to synchronise and demodulate. The suggested fix uses advanced pre-compensation methods. The user equipment (UE) uses its own location data and satellite ephemeris data to figure out the expected Doppler shift and make adjustments before sending data up.
• Long Propagation Delays: The round-trip time (RTT) to a LEO satellite can be tens of milliseconds, and hundreds of milliseconds for a GEO satellite. In terrestrial networks, the RTT is only microseconds. To make sure everything works right, the network’s timing advance mechanisms and protocols, like the Hybrid Automatic Repeat Request (HARQ), need to make up for these long delays.
• Seamless Handover Management: Keeping a user connected all the time while they move between the coverage areas of different fast-moving LEO satellites or between a satellite link and a terrestrial network (vertical handover) is a big problem. AI-driven handover decision algorithms are becoming more common in the solution space. These algorithms can predict where users are going and look at a lot of real-time metrics, like Received Signal Strength (RSS), Signal-to-Noise Ratio (SNR), and satellite elevation angle, to make proactive and best handover decisions. This keeps the service going and stops link failures.

To handle this complexity, the network core needs a new layer of abstraction. This is because of the AI-based automation discussed in the last section. This smart orchestration is what will make a bunch of separate networks on the ground, in the air, and in space work together as one smooth global system.

• The Convergence of Functions: Integrated Sensing and Communication (ISAC)

One of the most revolutionary architectural changes in 6G is the merging of two functions that have always been separate: wireless communication and environmental sensing. This paradigm, referred to as Integrated Sensing and Communication (ISAC) or Joint Communication and Sensing (JCAS), entails the collaborative design and utilisation of identical radio signals, spectrum, and hardware platforms to concurrently transmit data and sense the physical environment [42]. This is a big change in the network’s purpose; it will now be able to build a real-time digital twin of its physical surroundings instead of just sending data. The advanced multi-antenna systems we talked about earlier are what make ISAC work. For accurate sensing, Massive MIMO and XL-MIMO arrays must have a high spatial resolution [20]. These arrays can do advanced spatial beamforming, which means they can make multiple beams at the same time. Some of these beams are aimed at communication users to give them high data rates, and others are “sensing” beams that scan the area to find and follow objects.

This dual functionality, however, creates a fundamental performance trade-off that is at the heart of ISAC design. There are design goals that are at odds with each other in communication systems and sensing (radar) systems. Communication waveforms, such as Orthogonal Frequency-Division Multiplexing (OFDM), are designed to maximise data throughput and are often characterised by high peak-to-average power ratios and random-like structures, which are suboptimal for precise range and velocity estimation. On the other hand, traditional radar waveforms like linear frequency-modulated chirps are great for sensing because they have good correlation properties and constant envelopes. However, they are not good for embedding high-rate communication data. A primary research challenge in ISAC is the development of dual-functional waveforms and signal processing methodologies that can effectively or adaptively balance this trade-off according to application demands [43]. To make ISAC work well, you need more than just new ideas for the physical layer; you need a cross-layer architectural approach [44]. The physical layer creates the raw sensing measurements (i.e., reflected echoes), but this large amount of data needs to be moved, processed, and combined quickly, probably at the network edge, to meet latency limits. The resulting sensing information must then be exposed through well-defined interfaces to higher-layer network functions and external applications. This may necessitate the creation of new data planes specifically for sensing data and new network functions within the core network, such as a Sensing Management Function (SeMF), to orchestrate sensing tasks, manage resources, and resolve the trade-offs between the two functions. This deep integration makes ISAC a truly architectural transformation, not just a new physical layer feature.

The motivation for ISAC is multifaceted, offering significant benefits in efficiency, performance, and the creation of novel services. The core advantages include:

• Hardware and Spectrum Efficiency: By sharing the same RF front-end, antenna arrays, and signal processing units for both communication and sensing, ISAC significantly reduces hardware costs, device size, and power consumption compared to deploying two separate systems (e.g., a cellular base station and a radar system). Moreover, by using the same frequency spectrum for both functions, ISAC dramatically improves spectrum utilisation, which is critical as lower-frequency bands become increasingly congested.
• Enhanced Communication Performance: The sensing information gathered by the network is not just a secondary output; it can be used in a feedback loop to improve communication performance. For example, by sensing the location, orientation, and mobility of users and objects in the environment, the network can perform more accurate and proactive beamforming, predict link blockages, and reduce the overhead associated with channel state information (CSI) acquisition. This creates a symbiotic relationship where sensing improves communication, and communication provides the signals for sensing.
• Enablement of New Services and Applications: ISAC is a key enabler for a wide range of futuristic applications that require high-resolution environmental awareness. These include gesture and activity recognition for human–computer interaction, high-precision localisation and tracking for autonomous systems and indoor navigation, and environmental mapping and imaging for smart cities and disaster response. In essence, ISAC provides the foundational capability for the network to have a real-time, high-resolution understanding of the physical world, turning every 6G device into a potential sensing node.

• ISAC vs. Multi-Modal Sensing-Communication: An Evolutionary Perspective

While related, it is important to distinguish the concept of ISAC from the broader category of multimodal sensing-communication. The latter generally refers to systems where distinct, single-modal sensors (such as cameras, LiDAR, or GPS) are used to provide contextual information that aids the communication link. For example, a camera might detect an impending blockage, allowing the communication system to proactively switch to a different path.

ISAC represents a deeper, more fundamental level of integration. In an ISAC system, the communication signal itself is the primary sensing modality. The network does not rely on external sensors but rather analyses the echoes and distortions of its own transmitted waveforms to extract information about the environment (e.g., range, velocity, angle of objects). This can be thought of as an evolutionary step:

• Communication-Only Systems: The network’s sole purpose is data transmission.
• Sensing-Aided Communication: External, non-RF sensors provide data to optimise the communication link.
• Multi-Modal ISAC: The system may fuse information from its own radio-based sensing with data from other modalities (like cameras) to create a more comprehensive environmental picture.
• Truly Integrated Sensing and Communication (ISAC): The communication and sensing functions are co-designed and performed by the same signal and hardware, creating a unified framework.

This evolution from using sensing for communication to a system that performs sensing with communication is a cornerstone of the 6G architecture, enabling a proactive network that intelligently interacts with its physical surroundings.

3.2.4. Transforming Society: 6G Applications

The combination of advanced physical layer technologies and a new network architecture will not just be an academic exercise; it will lead to big changes in many areas of society and the economy. The transition to 6G is driven by the necessity to facilitate intricate, synchronised cyber-physical systems that depend on many-to-many communication with rigorous demands for latency, reliability, security, and environmental consciousness. This section examines the tangible impact of these developments through specific use cases, grounding the technological vision in empirical evidence from real-world trials and deployments [45,46,47].

• Digital Health and Immersive Medicine: The Continuum of Care

Next-generation wireless networks are set to create a continuum of healthcare that is more proactive, personalised, and accessible, moving care from the hospital to the home and the community. The ability of 5G, and eventually 6G, to reliably and securely support a massive number of connected devices is crucial for deploying wearable biosensors, smart implants, and other IoT health devices at a population scale. This enables the transition of Remote Patient Monitoring (RPM) from a niche service to a mainstream healthcare delivery model, allowing for continuous, real-time monitoring of patients with chronic conditions [48].

The connected ambulance is a great example of how URLLC can help people in real life. It becomes a mobile, real-time clinical hub. High-bandwidth, low-latency links let paramedics send high-definition video and a patient’s full vital signs to the emergency department of the hospital where they are going, giving staff important time to prepare. More importantly, it lets specialists who are far away be there in person to help paramedics with difficult tasks. It has been shown in practice that these kinds of ideas can work. For example, the Tekihealth trial in the UK successfully used 5G-connected diagnostic kits in care homes to do full remote check-ups, including otoscope and ECG readings, with the same level of accuracy as an in-person visit. This showed that high-quality remote care is possible.

Also, augmented and virtual reality (AR/VR) are moving from ideas to real-life use in medicine, thanks to advanced wireless networks that can handle a lot of data quickly. Some of the most important uses are immersive surgical training, remote expert collaboration, where a specialist can use an AR headset to “see” through the eyes of a surgeon in another location, and therapeutic uses, like using VR to manage pain.

• Intelligent Transportation: The C-V2X Cooperative Ecosystem

The development of wireless communication is closely tied to the idea of fully self-driving cars and safer, more efficient roads. Cellular Vehicle-to-Everything (C-V2X) technology, which is based on 5G and future 6G standards, is what makes it possible to create a cooperative transportation ecosystem where vehicles, infrastructure, and pedestrians all have real-time situational awareness [49]. C-V2X is a big step up from the older DSRC (Dedicated Short-Range Communications) standard. Its main benefits are much lower latency and higher reliability, which are both essential for safety-critical applications [26].

The performance advantages of C-V2X are not merely speculative; they have been empirically validated through comprehensive analysis. Traffic simulations on platforms such as SUMO have furnished substantial evidence of their effects [50]. One study showed that C-V2X can cut communication latency by more than 99% compared to DSRC. This directly means that safety will be better. The simulation showed that traffic conflicts would go down by 38% if 60% of cars were self-driving [26]. Real-world tests have confirmed these simulated results. In test environments like K-City in South Korea, self-driving cars have successfully used V2X communication to complete complicated tasks like navigating intersections and responding to changing traffic conditions. This proves that the technology is ready for use in real life. C-V2X connects vehicles to infrastructure (V2I) for traffic light and intersection data, to pedestrians (V2P) to protect vulnerable road users, and to the wider network (V2N) for traffic updates. This makes a complete information ecosystem. This constant, low-latency exchange of data makes it possible for cooperative manoeuvres like synchronised platooning and predictive collision avoidance. These are thought to greatly lower the number of traffic accidents and make things run more smoothly [26,51].

• Smart Cities and Digital Twins: The Sensing Fabric in Action

Massive IoT (mIoT) is made possible by the high connection density of 6G. It is the backbone of smart city projects that make services more efficient, sustainable, and enjoyable. The trend is moving away from just connecting individual assets and towards making integrated, data-driven systems that work like a digital twin of a city. The ISAC capability of 6G networks will be very helpful here. It will turn the communication infrastructure itself into a city-wide sensing fabric [52].

Smart waste management is a basic use case for smart cities. Putting sensors in public trash cans to keep an eye on how full they are is a clear improvement over a set collection schedule. But the real value comes out when this information is put into a bigger system. Cities can use this real-time data to change the routes for picking up trash in real time, which will save money on fuel, emissions, and operations. In a real-life example from Chicago, this data was even connected to predictive analytics tools to better manage pest control by predicting where overflowing bins would attract rodents. This shows a shift towards proactive, data-driven governance.

Smart lighting systems, on the other hand, do more than just save energy. Cities can keep an eye on air quality, hear gunshots for a faster police response, or offer public Wi-Fi by adding more sensors to connected LED streetlights. The lights can be changed on the fly in response to events happening right now, like making them brighter to help people find their way after a big event.

IoT sensors in smart parking and traffic management systems give real-time information on parking availability and traffic flow. This cuts down on traffic and emissions from drivers circling for a parking spot and helps city planners make better long-term decisions about infrastructure.

These apps show that we are clearly moving towards more complex, coordinated cyber-physical systems. They depend on many-to-many communication with strict standards for latency, reliability, and security. This has led to the creation of advanced network slicing and AI-driven resource management, which are key parts of the 6G architecture.

Table 6. Mapping 6G Applications to Network Requirements and Enabling Technologies.

Application Domain/Use Case	Key Network Requirements	Primary Enabling Technologies
Healthcare: AR-Assisted Remote Surgery	Latency: <1 ms (haptic feedback); Reliability: >99.99999% (seven nines); Bandwidth: >1 Gbps (HD/4K video); High Security and Privacy	URLLC, Mobile Edge Computing (MEC), High-Throughput XL-MIMO, Network Slicing, Zero-Trust Security
Transportation: Cooperative Collision Avoidance	Latency: <10 ms; Reliability: >99.999%; Positioning Accuracy: <10 cm; Sensing: Vehicle/VRU detection and tracking	C-V2X, Integrated Sensing and Communication (ISAC), Edge AI, High-Accuracy Positioning
Smart City: City-Scale Digital Twin	Connection Density: >107 devices/km2; Data Rate: High aggregate uplink; Sensing: Environmental monitoring; Coverage: Ubiquitous	Massive IoT (mMTC), Non-Terrestrial Networks (NTN), ISAC, AI-driven Analytics, Cloud/Edge Infrastructure
Immersive Media: Holographic Telepresence	Peak Data Rate: >1 Tbps; Latency: <10 ms (interaction); Jitter: Extremely low; Synchronisation: High precision	Terahertz (THz) Communications, XL-MIMO (Near-Field), Mobile Edge Computing (MEC), High-Throughput eMBB

Sources: Synthesised from [53].

shows how important application areas relate to their network needs and the main 6G technologies that make them possible.

3.2.5. Grand Challenges on the Path to 6G

The journey from the ambitious 6G vision to a globally deployed, operational reality is laden with significant challenges that span technology, security, sustainability, and societal trust. The technologies that make the 6G paradigm possible, like pervasive AI, extreme connectivity, and environmental sensing, also create new problems and risks that need to be dealt with from the ground up. Researchers, industry, and policymakers will all need to work together over the next ten years to overcome these challenges [54].

• The Trust Imperative: Security and Privacy in an Expanded Threat Landscape

The 6G network is thought of as an open, intelligent, and software-defined platform that will be deeply connected to all parts of society. But this deep integration makes the attack surface bigger and more complicated than ever, which brings about a new generation of security and privacy threats.

• Architectural Vulnerabilities: Open architectures like Open RAN are becoming more popular. They encourage flexibility and innovation, but they also make things more vulnerable. Open interfaces can let bad third-party code into the system, and, depending on a multi-vendor ecosystem, can lead to security implementations that are not always the same [16]. Also, because IoT deployments are so big, trillions of devices that could be cheap and not very secure could be used to access the network. These devices could be used to make botnets for big attacks like Distributed Denial of Service (DDoS) [55].
• AI-based Threats: The deep integration of AI creates a new type of threat that targets the network’s intelligence itself. Adversaries can use advanced adversarial machine learning (AML) methods, like data poisoning attacks to mess up the training data of network AI models or evasion attacks to trick AI-based intrusion detection systems. The AI that makes the network work better can be used as a weapon to plan smart, adaptable cyberattacks that are hard to find and stop [16,56].
• Quantum Threats: The long-term development timeline for 6G lines up with when quantum computing is expected to become more advanced. A sufficiently powerful quantum computer, employing Shor’s algorithm, could compromise public-key cryptographic algorithms (such as RSA and ECC) that support nearly all secure communications today, thereby rendering existing security protocols ineffective and jeopardising the long-term confidentiality of data [57].

Security must be a core principle of 6G design, not something that can be added later to deal with these many threats. One important way to fight back is to use a Zero-Trust Architecture (ZTA) [58]. ZTA is a big change from a security model that focuses on the perimeter to one that is based on the idea of “never trust, always verify.” In a ZTA, no user, device, or application is trusted by default, no matter where it is on the network. Every time someone tries to access a resource, they must be constantly and strictly authenticated, authorised, and encrypted. This method greatly lowers the chance that an attacker who has broken into one part of the system will move laterally. It is also necessary to protect the open, decentralised, and heterogeneous 6G environment [58]. Other important areas of research include creating and using Post-Quantum Cryptography (PQC) to protect against quantum threats, using AI to automatically find and respond to threats in real time, and looking into blockchain for decentralised identity and trust management [59].

• The Sustainability Paradox: Balancing Performance Growth with Energy Consumption

One of the main and clear goals of the 6G vision is to help build a digital society that lasts. But the network has a big problem with energy efficiency, which is often called the Jevons paradox or rebound effect [60]. While 6G technologies are being designed to use a lot less energy per bit, with a goal of 100 times better than 5G, the rapid growth in data traffic, the number of connected devices, and the huge computational load from AI could make the network and its data centres use a lot more energy [61]. This possible rebound effect, in which efficiency gains are outpaced by increased usage, presents a direct and formidable challenge to the goal of sustainability [62,63]. As wireless networks become increasingly pervasive and data traffic continues its exponential growth, their energy consumption has become a major global concern. The Radio Access Network (RAN) alone accounts for over 70% of a cellular network’s total energy usage. Consequently, sustainability, and specifically energy efficiency, has moved from a secondary consideration to a primary design goal for 6G, enshrined as a core KVI.

To solve this problem, we need a “Green 6G” strategy that looks at the whole lifecycle and operation of the network, not just hardware improvements. The “Green 6G” vision encompasses two main objectives: minimising the direct energy consumption of the network itself and enabling sustainability in other sectors of the economy through 6G-powered applications.

• Technology-Driven Savings: This means creating and using naturally energy-efficient technologies. For instance, RIS can lower the transmit power needed by base stations and UEs by using passive elements to improve link quality. Research into energy harvesting (EH) techniques seeks to empower low-power devices to extract energy from ambient RF signals, thereby diminishing their dependence on batteries.
• Architectural and Operational Savings: A lot of the energy used by a network is used by devices that are “always on” but not sending data. “Less ON, More OFF” is one of the design ideas behind 6G. It aims to make network components, from single antennas to whole base stations, sleep aggressively and smartly, waking them up only when they are needed. AI will be very important here because it will be able to predict traffic patterns to make the best use of resources and turn off parts of the system before they are needed.
• User-Centric and Economic Approaches: A new idea for Green 6G is to make energy use visible and useful for both users and applications. This could mean showing users information about the carbon footprint and energy cost of different services so they can make “green” choices, like choosing a lower video resolution when high fidelity is not needed. This could also lead to new ways of conducting business, like giving users “energy credits” for choosing services that have less of an impact.
• Lifecycle Sustainability: Real sustainability goes beyond how much energy is used in operations. It also needs to think about how making, using, and throwing away network equipment affects the environment. It is very important to follow the rules of the circular economy, such as making hardware that lasts, can be upgraded, and can be recycled. Architectural trends such as Open RAN, which advocate for softwarisation and disaggregation, can also play a role by diminishing the necessity to replace specialised hardware with each upgrade cycle [64].

• The Path to Realisation: Open Problems and Future Directions

The path from the 6G vision to a deployed, working reality is long and full of basic open questions that will shape the wireless research agenda for the next ten years. Combining ideas from forward-looking workshops and literature, we can see that there are several important research needs that are necessary to handle the unprecedented size and complexity of the 6G system.

• New Network Architectures: The internet’s current layered architecture is very successful, but it may not be flexible or efficient enough for 6G. A major research challenge is to create a new, more flexible architecture that can easily combine communication, computation, sensing, and storage into one system. This architecture needs to be able to handle the huge complexity of a hybrid terrestrial-NTN, multi-band system and offer performance guarantees that can be verified for a wide range of services.
• Hardware and Spectrum Frontiers: It is still important to push the limits of technology. This entails ongoing investigation into novel semiconductor materials (e.g., III-V compounds, wide-bandgap semiconductors) and photonic technologies to develop practical, economical, and robust components for the THz bands [20]. At the same time, it is important to create more advanced dynamic spectrum access and sharing methods so that all frequency bands from sub-6 GHz to THz can be used effectively, especially when terrestrial and non-terrestrial networks need to work together [18].
• The Convergence of AI and Control Theory: AI provides a potent data-driven methodology for network management; however, its “black box” characteristic may occasionally lack the formal assurances of stability and robustness offered by conventional model-based control and optimisation theory. One important goal for the future is to create a unified theoretical framework for real-time network control that combines the learning and adaptation abilities of AI/ML with the analytical rigour of classical control theory. This will make autonomous networks more stable and predictable [65].
• Managing Scalability and Complexity: The 6G vision is huge: trillions of connected devices, antenna arrays with tens of thousands of elements, and AI-driven control loops that work in microseconds. It is very important to deal with the basic problems of scalability. This entails the creation of novel algorithms, protocols, and management systems capable of operating efficiently and resiliently at this unparalleled scale, without succumbing to computational or communication overhead, thereby guaranteeing that the network’s management traffic does not deplete the resources it is intended to supply [66].

4. Discussion

4.1. Summary of Evidence

This scoping review mapped the extensive literature defining the transition to 6G wireless systems. The evidence synthesised reveals a consensus that 6G is not an incremental upgrade but a paradigm shift. The key findings are threefold. First, the vision is driven by a move from technology-centric Key Performance Indicators (KPIs) to value-centric Key Value Indicators (KVIs), emphasising sustainability, trustworthiness, and digital inclusion. Second, this vision is enabled by a triad of interdependent physical layer technologies—RIS, THz communications, and XL-MIMO—that collectively engineer a smart radio environment. Third, these technologies necessitate an architectural overhaul towards a self-orchestrating, AI-native framework that unifies terrestrial and non-terrestrial networks and converges the functions of communication and sensing.

4.2. Limitations

This review has several limitations. The search was restricted to English-language publications, potentially omitting relevant research from other regions. The rapidly evolving nature of 6G research means that new developments may have emerged since the search was conducted. Furthermore, as a scoping review, we did not perform a critical appraisal of the quality of the included sources of evidence; the goal was to map the breadth of the field rather than assess the validity of individual studies.

4.3. Conclusions

The shift from 5G to 6G is not just a small change in technology; it is a complete rethinking of the role of wireless networks in society. It signifies a transformative evolution from a mere communication utility to an intelligent, multifunctional cyber-physical framework engineered for sustainability, reliability, and universal accessibility. A complicated mix of revolutionary physical layer technologies, a complete change in architecture, and a new definition of performance, going from technology-centred metrics to value-centred societal goals, is all driving this change.

This survey has provided a comprehensive analysis of this transition, guided by the fundamental paradigm shift from technology-centric KPIs to value-centric KVIs. Our critical examination of the foundational physical layer technologies—RIS, THz, and XL-MIMO—has revealed not only their individual promise but also their crucial synergistic interplay. The comparative analyses of RIS versus active relays and XL-MIMO versus Cell-Free architectures have illuminated the key trade-offs in performance, cost, and efficiency that will shape 6G network design.

Many problems need to be solved before 6G can happen, but they can all be grouped into two main problems: making a network that people can trust and one that will last. The goal of 6G is to become the smart nervous system of a society that is becoming more digital. But the technologies that make this vision possible—widespread AI, huge amounts of data collection from integrated sensing, and hyper-connection across open, decentralised architectures—also bring new risks to security, privacy, and trust in society. There is a “trust paradox” here: the more powerful and connected the network becomes, the more damage it could do if it is hacked. So, using basic ideas like Zero-Trust Architecture to solve security problems is not only a technical need, but it also means building a network that society can trust to do its most important tasks.

At the same time, the size of the 6G vision is a direct threat to the environment. The expected rapid rise in network traffic, connected devices, and the huge processing power needed for pervasive AI is leading to a huge use of energy and resources. The environmental cost of this new connectivity could outweigh its social benefits if there is not a proactive and comprehensive “Green 6G” strategy that deals with the energy efficiency paradox.

These two big problems are linked in a way that cannot be broken. It takes a lot of energy to run advanced AI-driven security and post-quantum cryptography. On the other hand, if not carefully planned, aggressive energy-saving measures could hurt the resilience of the network and the ability to monitor security. Consequently, forthcoming research must traverse the intricate, multifaceted trade-offs among these objectives. Ultimately, the breakthroughs that will define 6G will be measured not only by performance metrics like terabits per second, but also by how they help create a system that is safe, private, strong, and energy-efficient all at the same time. This all-encompassing view must shape the next ten years of research, standardisation, and policymaking to make sure that 6G lives up to its big promise of a world that is better connected, smarter, and more sustainable.

The transition from 5G to 6G is not merely an incremental technological step but a fundamental reimagining of the role of wireless networks in society. It represents a paradigm shift from a pure communication utility to an intelligent, multi-purpose cyber-physical fabric designed to be sustainable, trustworthy, and ubiquitously available. This transformation is driven by a complex interplay of revolutionary physical layer technologies, a complete architectural metamorphosis, and a redefinition of performance itself, moving from technology-centric metrics to value-centric societal goals.