Smart Networks and Services Joint Undertaking (SNS JU)

6G Architecture Working Group

Pre-Release Version for Public Consultation

10 March 2025

WHITE PAPER
TOWARDS 6G ARCHITECTURE:
KEY CONCEPTS, CHALLENGES,
AND BUILDING BLOCKS

DOI: 10.5281/zenodo.15001377
URL: https://doi.org/10.5281/zenodo.15001377

Date:
6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

STATEMENT ON PRE-RELEASE VERSION

The current pre-release version of the white paper is aimed for the public consultation. Thus,
upon collection of feedback, various edits and modifications can be applied towards the final
version of the white paper.

Dissemination level: Public | 2

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

ACKNOWLEDGEMENT
Co-funded by the European Union. Views and opinions expressed are however those of the
author(s) only and do not necessarily reflect those of the European Union or the SNS-JU. Neither
the European Union nor or the SNS-JU can be held responsible for them.

Dissemination level: Public | 3

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

EXECUTIVE SUMMARY
We are entering the standardization phase for the 6th generation (6G) of wireless
technologies. While valuable lessons have been learned from the design, deployment, and
operation of 5G and Beyond in Europe, new requirements, emerging technologies, and evolving
business models must be natively integrated into the next-generation mobile network
architecture.

This white paper presents a comprehensive snapshot of the current architectural

considerations explored by the Smart Networks and Services Joint Undertaking (SNS-JU) projects.
It aims to discuss the rationale for novel architectural components, the ongoing design efforts,
and the future outlook for 6G.

We analyse the blueprint for next-generation mobile networks, building on past experiences
while integrating cutting-edge advancements. The structure follows the IMT-2030 framework,
categorizing insights into the key usage scenarios and overarching architectural aspects.

• Usage Scenarios: We analyse the evolution of IMT-2030 paradigms, focusing on

Immersive, Massive, and Hyper-Reliable Low-Latency Communications. We also explore
novel 6G enablers, such as Ubiquitous Connectivity, AI-driven Communication, and
Integrated Sensing & Communication.
• Overarching Aspects: The next-generation architecture must natively embed Security,
Privacy & Trustworthiness, Sustainability, and Network Exposure Capabilities—ensuring
these critical aspects are foundationally considered rather than retroactively
incorporated.

Finally, we conclude by outlining the essential building blocks shaping the next-generation
mobile network architecture, highlighting the most promising research paths identified in this
white paper.

Dissemination level: Public | 4

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

TABLE OF CONTENTS
STATEMENT ON PRE-RELEASE VERSION ............................................................................................ 2
ACKNOWLEDGEMENT ........................................................................................................................ 3
EXECUTIVE SUMMARY ....................................................................................................................... 4
1 INTRODUCTION............................................................................................................................. 8
1.1 Current Deployments ........................................................................................................................... 9
1.1.1 Current Deployments and Efforts ............................................................................................... 9
1.1.2 Main Use cases and Technology Gaps ...................................................................................... 12
1.2 References .............................................................................................................................................. 13
2 SYSTEM BLUEPRINT .................................................................................................................... 14
2.1 Architectural Design Principles..................................................................................................... 14
2.2 System blueprint ................................................................................................................................. 16
2.3 Migration aspects ................................................................................................................................ 19
2.4 Modular Architecture Design ......................................................................................................... 20
2.5 References .............................................................................................................................................. 22
3 EXTENSIONS ................................................................................................................................ 23
3.1 Integration of Overarching Concepts into the 6G Architecture ....................................... 23
3.1.1 Resource Management in the Deep-Edge-Edge-Cloud Continuum.......................................... 23
3.1.2 Sustainability and compute Continuum.................................................................................... 26
3.1.3 Zero-Trust Security and Interoperability .................................................................................. 28
3.1.4 Sub-Network Integration in 6G Networks ................................................................................ 30
3.2 Integration of 6G Network Paradigms........................................................................................ 31
3.2.1 Network Slicing and Multi-Access Edge Computing (MEC)....................................................... 31
3.2.2 Time-Critical and Deterministic Communications .................................................................... 35
3.3 Towards a Global SBA ....................................................................................................................... 37
3.3.1 Global Service Based Architecture ............................................................................................ 37
3.3.2 Bringing The GSBA to the RAN .................................................................................................. 37
3.4 References .............................................................................................................................................. 39
4 UBIQUITOUS NETWORKS............................................................................................................ 40
4.1 Ubiquitous Coverage via 6G NTN Architecture ...................................................................... 40
4.1.1 Distribution of core network between terrestrial and non-terrestrial layers........................... 42
4.1.2 Unified MANO ........................................................................................................................... 44
4.1.3 E2E control of UPF based on multi-domain SDN ...................................................................... 47
4.1.4 Direct handheld access for UE .................................................................................................. 48
4.1.5 Distributed NG-RAN for 5G/6G Unified NTN Networks ............................................................ 49
4.2 Multi-Connectivity for 6G Ubiquitous Coverage .................................................................... 52
4.2.1 6G Multiconnectivity ................................................................................................................ 52
4.2.2 NTN Integration for 6G Multiconnectivity ................................................................................ 54
4.3 Confluent Transport Network ....................................................................................................... 55
4.4 References .............................................................................................................................................. 58
5 ARTIFICIAL INTELLIGENCE AND COMMUNICATIONS .................................................................. 60

Dissemination level: Public | 5

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

5.1 AI/ML Framework / Intelligent Plane ........................................................................................ 60

5.1.1 AI/ML enablers and framework ................................................................................................ 61
5.1.2 Intelligence Plane in O-RAN ...................................................................................................... 62
5.1.3 Cloud Native Data Plane ........................................................................................................... 63
5.1.4 AI Air Interface .......................................................................................................................... 65
5.2 Intent-/Goals-Driven Communications ..................................................................................... 67
5.2.1 Intent-driven Cognitive 6G network ......................................................................................... 67
5.2.2 Goal-Oriented and Semantic Communication in 6G AI-Native Networks ................................ 68
5.2.3 6G ML training, Intent based interface, Network Digital Twin ................................................. 70
5.3 Management and Orchestration ................................................................................................... 74
5.3.1 Edge Control ............................................................................................................................. 74
5.3.2 6G Network Management and Automation ............................................................................. 75
5.3.3 Native AI - pervasive monitoring system .................................................................................. 77
5.4 Digital twin............................................................................................................................................. 79
5.4.1 Integrating Network Digital Twins into 6G Architectures ......................................................... 79
5.4.2 AI-Driven and MLOps-Enabled Network Digital Twin for Multi-Domain Communications ...... 81
5.5 References .............................................................................................................................................. 83
6 INTEGRATED SENSING AND COMMUNICATION ......................................................................... 87
6.1 Integration of non-3GPP and 3GPP Sensing ............................................................................. 88
6.2 Distributed Sensing Architecture ................................................................................................. 89
6.3 Optical Wireless Communication based ISAC ......................................................................... 92
6.4 Sensing Service Provisioning and Exposure ............................................................................ 94
6.5 ISAC Application for V2X Use cases ............................................................................................. 97
6.6 MultiX -Advancing ISAC through multi-technology, multi-sensor fusion, multi-band
and multi-static perception................................................................................................................................. 100
6.7 References ............................................................................................................................................ 101
7 SECURITY, RESILIENCY, PRIVACY AND TRUSTWORTHINESS ..................................................... 103
7.1 Security concerns introduced by 6G ......................................................................................... 103
7.2 Trustworthiness as the pilar to build safe, resilient & reliable security and privacy
solutions ................................................................................................................................................................... 105
7.2.1 Trustworthiness and Level of Trust Relation .......................................................................... 106
7.2.2 6G Properties in Trusworthiness............................................................................................. 107
7.3 Challenges ............................................................................................................................................ 109
7.4 Overarching blocks ........................................................................................................................... 116
7.4.1 Trustworthiness ...................................................................................................................... 117
7.4.2 Design layer ............................................................................................................................ 119
7.4.3 Operation Layer ...................................................................................................................... 119
7.4.4 AI/ML Layer ............................................................................................................................. 120
7.4.5 Sensing .................................................................................................................................... 122
7.5 References ............................................................................................................................................ 122
8 SUSTAINABILITY ........................................................................................................................ 124
8.1 Introduction ........................................................................................................................................ 124
8.2 RAN sustainability advancements.............................................................................................. 127
8.2.1 relay nodes for energy-efficient ran ....................................................................................... 127

Dissemination level: Public | 6

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

8.3 Transport Network sustainability advancements ............................................................... 130

8.3.1 Optical transport networks supporting sustainable capacity scaling ..................................... 130
8.3.2 Optical transport networks supporting flexible functional splits ........................................... 133
8.4 E2E Sustainability Advancements .............................................................................................. 135
8.4.1 Real-time sustainable resource allocation in integrated TN-NTNs ......................................... 135
8.5 Conclusions .......................................................................................................................................... 138
8.6 References ............................................................................................................................................ 139
9 NETWORK EXPOSURE CAPABILITIES ......................................................................................... 141
9.1 Network Exposure capabilities ................................................................................................... 141
9.1.1 Enabling deterministic networking by efficiently bridging multiple network domains .......... 141
9.1.2 API ecosystem for exposure and interconnection services .................................................... 143
9.1.3 Exposure services to enable advanced extended reality applications.................................... 144
9.1.4 Exposure services to enable Connected and Automated Mobility (CAM) .............................. 145
9.2 Programmability enabling features ........................................................................................... 146
9.2.1 Infrastructure management layer ........................................................................................... 146
9.2.2 Programmable transceivers in optical Transport networks .................................................... 147
9.3 Network Representation and Functional structure ............................................................ 147
9.3.1 Network digital twin ............................................................................................................... 147
9.3.2 Network abstraction to support trials .................................................................................... 148
9.3.3 User centric network structuring ............................................................................................ 150
9.4 References ............................................................................................................................................ 151
10 6G ARCHITECTURAL DEFINITION .............................................................................................. 152
10.1 Modular Architecture Design ....................................................................................................... 152
10.1.1 Multi Layered Architectures for NTN Integration ............................................................... 152
10.1.2 Time Critical and Deterministic Networking Integration .................................................... 153
10.1.3 Integration of Sensing and Digital Twinning ....................................................................... 155
10.1.4 Sustainability Management ................................................................................................ 156
10.1.5 Global Service Based Architecture ...................................................................................... 157
10.2 Cloud Continuum Management ................................................................................................... 158
10.3 Interoperability and Global Operation ..................................................................................... 159
10.4 AI driven Network Management and Orchestration .......................................................... 160
10.5 References ............................................................................................................................................ 161
11 ABBREVIATIONS AND ACRONYMS ............................................................................................ 162
12 LIST OF EDITORS & REVIEWERS ................................................................................................ 170
13 LIST OF CONTRIBUTORS ............................................................................................................ 172
SUPPORTING PROJECTS ................................................................................................................. 177

Dissemination level: Public | 7

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

1 INTRODUCTION
As the telecommunications industry advances beyond 5G, the transition to 6G is set to
revolutionize the way networks are designed, deployed, and utilized. The 6G Architecture
Working Group has prepared this white paper to define the fundamental architectural principles
that will guide the development of next-generation mobile networks. The white paper provides
an extensive analysis of the key technological enablers, system design choices, and research
challenges that will shape the 6G ecosystem.

Unlike previous generations, 6G networks will move beyond connectivity to become

intelligent, context-aware, and adaptive systems, leveraging artificial intelligence (AI), deep-edge
computing, non-terrestrial networks (NTN), and integrated sensing and communication (ISAC).
This transition will require a fundamental rethinking of network architecture, ensuring that 6G
can support ubiquitous, sustainable, and resilient communication across heterogeneous
environments, from urban landscapes to remote and un(der)served areas.

This white paper begins by analysing the state of current 5G and Beyond 5G (B5G)
deployments, highlighting experimental testbeds and federated research platforms across
Europe that integrate cutting-edge technologies like network slicing, AI-enabled orchestration,
multi-access edge computing (MEC), real-time extended reality (XR) applications, and NTN
testbeds. Despite these advancements, critical technology gaps persist in latency, energy
efficiency, network scalability, and interoperability with respect to the demanding requirements
of existing and future vertical use cases. Key challenges include latency constraints for time-
sensitive applications, throughput limitations for future applications like holographic
communication, energy efficiency concerns for sustainability of networks, and scalability and
interoperability issues. This sets the stage for the 6G system blueprint, detailing how 6G will
address these challenges.

The 6G System Blueprint outlines the key architectural principles that define 6G’s end-to-end
design. These principles include cloud-native, AI-driven networks for dynamic, intelligent
network management; modular and scalable architectures to support diverse use cases; intent-
based networking (IBN) for dynamic adaptation to user intent; ISAC for real-time environment
perception. The 6G architecture will be built upon a multi-layered framework consisting of an
infrastructure layer integrating terrestrial, aerial, and space-based network components and
resources; a network layer with a unified, software-defined design incorporating 6G radio access
networks (RANs), core network functions, and beyond-communication services enablers; an
application layer with an AI-driven service framework for customized network functionality; and

Dissemination level: Public | 8

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

a security and trust layer with a decentralized zero-trust security model. This blueprint redefines
network flexibility, enabling adaptive, programmable, and AI-powered connectivity.

Key enhancements will shape the evolution of 6G, including resource management in the deep-
edge-edge-cloud continuum, leveraging distributed computing and AI-driven workload
orchestration; zero-trust security and interoperability through zero-trust networking (ZTN) and
standardized APIs; and sub-network integration in 6G networks, featuring dynamic, software-
defined sub-networks. 6G will extend connectivity beyond terrestrial networks by NTNs for
global coverage, enabling direct handheld access to satellites, and incorporating optical wireless
communications (OWC) for seamless indoor and outdoor coverage. AI will be native to 6G,
embedding intelligence into network automation, intent-based communications, and digital
twins for predictive analytics. 6G will merge wireless communication with environmental
sensing, enabling smart cities, industrial IoT, vehicular communication, and healthcare
innovations.

Security, resilience, and trust in 6G will be ensured through a multi-layered security

framework covering end-to-end encryption, post-quantum cryptography, AI-driven threat
detection, and resilient network architectures. Sustainability is a core pillar of 6G, with energy-
efficient architectures addressing green AI-powered network management, sustainable
hardware design, and intelligent resource allocation. Finally, the white summarizes the main
findings on the architectural components, highlighting trends, solutions, and highlighting a
roadmap towards the 6G architectural design. This white paper serves as a comprehensive guide
to 6G architectural advancements, design principles, and emerging technologies, aiming to
redefine global connectivity by addressing flexibility, intelligence, security, and sustainability.

1.1 CURRENT DEPLOYMENTS

1.1.1 CURRENT DEPLOYMENTS AND EFFORTS

In the context of the SNS projects ecosystem, there is a continued effort in deploying federated,
sustainable experimentation platforms that span across Europe, enabling advanced research and
validation of 5G and Beyond 5G (B5G) technologies. They provide decentralized and scalable
architecture supporting also Testing as a Service (TaaS), resource sharing, and cross-domain
application deployment. Some of them also support Portals that serve as the primary entry point,
offering tools for experimenters, infrastructure owners, and vertical developers.

The testbed deployments integrate cutting-edge technologies such as network slicing, AI-
enabled components, and real-time and intelligent orchestration, supporting diverse use cases,
including edge computing, immersive applications, and data-intensive tasks. These testbeds also

Dissemination level: Public | 9

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

facilitate rapid prototyping, KPI tracking, and interoperability testing, bridging the gap between
experimental platforms and commercial-grade networks.

Table 1.1 provides a comprehensive summary of the testbeds, their capabilities, and key
features.

Table 1.1: Current technology deployment in the SNS projects ecosystem

Testbed Key Innovations and

Location Deployment Details Infrastructure
Name Additional Features

Greek Patras, Private 5G network with 1082 CPUs, 4.5 TB AI-enabled network
Testbed Greece Network Slice as a Service RAM, 100 TB configuration, tailored
(Patras) (NSaaS) model. Indoor: 4 storage, NVIDIA network slices, open-
picocells on FR1 (band n78). GPUs, source OSS for
Outdoor: 2 microcells on 100GbE/400GbE experimentation,
FR1. switches, cloud hosting with
Kubernetes clusters. MEC, and pre-
provisioned SIMs for
external devices.

Spanish Malaga, Outdoor: 4 microcells on FR1 668 cores, 2.6 TB Includes downtown
Testbed Spain (band n78), 2 macro cells on RAM, 13 TB shared Malaga, Harbor, and
(Malaga) mmWave FR2 (n257, n261). storage, NVIDIA Torremolinos via
[SANDBO Indoor: 2 picocells on FR1. GPUs, OpenNebula MOCN. Real-time
XD2.1] for infrastructure orchestration, AI-
management. assisted monitoring,
Open5GS Release-17
compliance, and
immersive
applications for XR
experiences.

Norwegia Oslo, Multi-vendor 5G SA core, Commercial-grade Zero-Touch Service

n Testbed Trondhei RAN coverage at multiple and experimental Management (ZSM),
[FUDGED m, geographic locations, components, Nokia dynamic
3.2] Svalbard "Network on Wheels" for Orchestration orchestration,
remote coverage. Center (NOrC) for adaptive traffic
end-to-end aggregation, mobile
orchestration. "NOW" platform for
flexible
experimentation, and
seamless
commercial/experime
ntal integration.

Italian Turin, Commercial 5G NSA network Advanced data Focused on real-time

Testbed Italy (Rel-15, 3.7 GHz), dark fiber processing XR experiences with
(Turin) connections. capabilities for XR, advanced network
[TrialsNe immersive slicing and traffic
tD2.3] applications, management, 10 GbE
transitioning to Rel- dark fiber for high-
16/17. speed data transfer,
and enhanced low-
latency deployments.

Dissemination level: Public | 10

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Romania Bucharest Dual 5G NSA/SA architecture Edge computing AI-enhanced tools for
n Testbed , Iasi with local breakout for edge facilities integrated KPI tracking, federated
[TrialsNe computing. with federated edge deployments,
tD2.3] IPFABRIC. support for scalable
and adaptive
experiments, and low-
latency edge
processing for time-
sensitive applications.

Spanish Madrid, 5G SA network with mid- Experimental setups Supports applications

Testbed Spain band and mmWave for vertical-specific in logistics, media, and
(Madrid) capabilities. use cases, advanced immersive
[TrialsNe orchestration technologies,
tD2.3] frameworks. leveraging mid-band
and mmWave
configurations for high
throughput and low
latency experiments.

SUNRISE- Pan- Federation of 6G test Decentralized multi- AI-enabled cloud

6G European infrastructures with cross- domain testbed with continuum, cross-
Testbed (Multiple domain application heterogeneous testbed orchestration,
Locations onboarding, testing as a computing federated access to 6G
) service, and a unified resources, enablers, and real-time
experimentation framework. intelligent metaverse
orchestration, and applications with
Camara-based HoloMIT for
standardized APIs. volumetric
holography.

6G-XR Barcelona XR-focused experimentation Edge-cloud Focused on enabling

Testbed , Spain platform for immersive continuum real-time volumetric
(I2CAT, applications and AI-driven resources with XR applications, QoE-
Spain) optimization. adaptive driven resource
orchestration for allocation, and
real-time XR integration with
rendering. federated 6G
platforms.

6G-Bricks Distribut Cloud-native infrastructure Federated resources Supports AI-based

Testbed ed for AI-driven with dynamic intent-driven
(EUR, Locations experimentation in 6G workload balancing networking, low-
Europe- environments. and optimized latency XR
wide) cloud-edge applications, and
integration. integration with other
European testbeds.

This federated experimentation platform exemplifies a significant leap in telecommunications

research, offering a cohesive ecosystem that bridges diverse testbeds across Europe. With its
decentralized architecture, advanced tools, and shared resources, the initiative fosters innovation
in 5G and Beyond 5G (B5G) technologies, empowering researchers and industry stakeholders to
explore emerging use cases and validate cutting-edge solutions.

Dissemination level: Public | 11

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

The unique capabilities of each testbed, ranging from AI-enabled orchestration to real-time XR
deployments and edge computing, demonstrate the potential for scalable, adaptive, and
collaborative research. By aligning with global standards and supporting interoperability through
unified APIs, the platform ensures broad accessibility and long-term sustainability.

As the industry transitions towards 6G, this initiative lays a strong foundation for collaborative
innovation, addressing the challenges of future networks while driving economic growth and
technological advancement in Europe and beyond.

1.1.2 MAIN USE CASES AND TECHNOLOGY GAPS

The experimental facilities discussed above are used to deploy a range of compelling vertical
use cases involving high-demand requirements, which are facing gaps in key areas:

• Latency Requirements: Applications like "Smart Crowd Monitoring" and "Remote

Proctoring” demand ultra-low latency (<10ms) for real-time operation. Current 5G NSA
setups struggle to achieve these thresholds, with upgrades to 5G SA and new network
releases (e.g., Rel-17) anticipated to bridge this gap.

• Throughput Limitations: Use cases involving XR experiences or data-heavy applications

in healthcare face challenges in maintaining consistent uplink and downlink throughput,
particularly during peak traffic periods. Despite incorporating mid-band and mmWave
frequencies, further capacity enhancements are necessary [Fer+22].

• Coverage and Reliability: Mixed urban and rural environments present challenges for
achieving consistent service quality. Local breakout and edge computing solutions
mitigate some issues, but gaps persist, especially in regions reliant on 5G NSA
infrastructure.

• Energy Efficiency: Use cases requiring extensive connectivity (e.g., IoT networks) often
struggle with power consumption optimization. Intelligent traffic management and
energy-efficient network strategies are crucial for addressing this issue.

High-demand scenarios like live media streaming and uplink-intensive use cases face unique
complexities. These include dynamic resource allocation, seamless integration across diverse
networks, and programmatic E2E orchestration. These challenges include:

• Dynamic Resource Allocation: Meeting fluctuating bandwidth needs for uplink-heavy

applications while balancing QoS and QoE targets [Mon+22].

• Latency and QoS Guarantees: Managing stringent latency and reliability requirements in
shared environments.

Dissemination level: Public | 12

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

• Interoperability: Ensuring seamless integration across multi-vendor equipment and

proprietary systems for E2E orchestration.

• Advances in orchestration frameworks and adherence to standards like 3GPP and

GSMA/Camara APIs will enable more effective traffic isolation and management for
uplink-heavy use cases [TrialsNetD2.3].

Federated experimental platforms require further development to support advanced

interoperation and vertical integration. Key gaps include:

• Horizontal and Vertical Interoperation: Platforms must enhance cross-domain

coordination to support scalable and dynamic experimentation.

• Service Reliability and Energy Efficiency: Ensuring consistent service quality and power-
efficient operations across diverse use case scenarios.

• Network Upgrades: Broad adoption of 5G SA, improved edge computing, and enhanced
spectrum usage are critical to addressing latency, throughput, and coverage challenges.

While current platforms demonstrate promising results, gaps in performance, scalability, and
resource efficiency must be addressed to fully realize their potential. Integrating emerging
solutions such as the ones that we propose in this will play an important role for efficiently
integrating this view.

Advancing experimental platforms and leveraging collaborative research across Europe are
essential for enabling next-generation communication services. These efforts will drive
innovation while ensuring sustainability, adaptability, and interoperability in the evolving 5G and
Beyond 5G ecosystem.

1.2 REFERENCES
[FUDGED3.2] P. Chakraborty and M.-I. Corici, “FUDGE-5G D3.2 On-boarding and Deploying of the Vertical
Use Cases”, 2024.
[Fer+22] S. Fernández, M. et al., “Multiparty Holomeetings: Toward a New Era of Low-Cost Volumetric
Holographic Meetings in Virtual Reality”, IEEE Access 2022.
[SANDBOXD2.1] G. Makropoulos, “Deliverable D2.1 - Ecosystem analysis and 6G-SANDBOX facility design”,
Zenodo, Jul. 2023. doi: 10.5281/zenodo.8366395.
[Mon+22] M. Montagud, et al., “Towards SocialVR: Evaluating a Novel Technology for Watching Videos
Together”, Virtual Reality (Springer), 2022.
[TrialsNetD2.3] G. Scivoletto, “Deliverable D2.3 - Final design of Platforms and Networks solutions”,
Zenodo, Dec. 2024. doi: 10.5281/zenodo.14512906.

Dissemination level: Public | 13

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

2 SYSTEM BLUEPRINT
There is a need for a more flexible and adaptable system architecture for 6G E2E compared to
previous generations. 6G will need to support an efficient mobile communication service as well
as the new 6G offerings such as beyond communication services, AI and compute offloading.
Further on, 6G will also be able to process data, generate insights, and deliver additional valuable
services such as spatial/temporal data services, compute services to applications, and intelligent
services (e.g., such as AI functionality, analysis, and optimizations).

Additionally, a wider ecosystem collaboration as compared to the previous generations,

including mobile network operators, cloud providers, enterprises, vertical industries, integrators,
application developers, application service providers, end users, etc., will be a prerequisite to
creating values for all the ecosystem players. This needs 6G networks to be a versatile platform
interfacing with a broad range of applications and that can be tailored to all the ecosystem players
to support their specific needs.

The architecture should also support the ability to better scale networks than in today’s 5G
deployments. This should be done dynamically, based on current needs, to improve the efficiency.

2.1 ARCHITECTURAL DESIGN PRINCIPLES

There is a need for a new flexible and adaptable 6G E2E system architecture, supporting the
efficient next generation mobile communication and the new beyond communication services as
well as the AI and security elements. This need maps to any use case.

The 6G platform will expose a diverse set of Application Programming Interfaces (APIs) to
applications, users, and industry verticals, with the capability to extend these APIs over time.
Services provided by the 6G platform will be implemented through the 6G end-to-end (E2E)
system, which comprises multiple interacting subsystems. These subsystems encompass device
evolution, network infrastructure enhancements, advanced network capabilities, and pervasive
functionalities including security and privacy, data handling, artificial intelligence frameworks,
and end-to-end management and orchestration. The 6G E2E system will adhere to the ten
architectural principles detailed in Table 2.1, offering an efficient framework to support emerging
6G technologies.

Dissemination level: Public | 14

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Table 2.1: Mapping of architecture design principles and their impact of 6G E2E system design

Design Key values

Impact on 6G E2E system blueprint design
principles

Support and This feature encompasses generic and dynamic exposure Economic
exposure of 6G functionalities, e.g. simplified APIs to expose capabilities to E2E sustainability,
services and applications facilitating seamless integration of beyond environmental
capabilities communication network functions and hardware capabilities. It sustainability,
emphasizes the inclusion of pervasive AI and a robust compute Trustworthiness,
infrastructure to enhance the overall flexibility for compute Inclusiveness
offloading of services and improve the performance of the system.

Full The 6G E2E system necessitates a comprehensive and widespread Economic

automation data and analysis framework, complemented by a pervasive AI sustainability,
and framework and service management and orchestration. The environmental
optimization emphasis is on building a pervasive infrastructure that enables sustainability,
efficient data handling, robust AI integration, and seamless service Trustworthiness
orchestration across diverse scenarios. The pervasive AI
framework provides means for predictive orchestration,
distributed AI/ML agents to optimize the system without human
interaction. It provides continuous orchestration over multiple
administrative domains, supporting the multi-stakeholder 6G
ecosystem.

Flexibility to This design principle is for incorporating many different Economic

different modalities of connectivity, including local and wide area networks, sustainability,
network non-terrestrial networks, subnetworks, public and private environmental
scenarios networks with a network architecture that supports seamless sustainability,
mobility between them. It is for utilizing new spectrum in an Inclusiveness
efficient way, implementing programmable transport
configurations and having application awareness and adaptive
quality of service (QoS) and quality of experience (QoE).

Scalability This focuses on creating a pervasive service management and Economic

orchestration system (e.g., scaling up and down based on mobility sustainability,
and time-varying traffic needs), incorporating a network-centric environmental
exposure layer and optimized transport network functions (e.g., sustainability,
over heterogeneous multi-domain/multi-clouds). In addition, Trustworthiness
network modularity will enable dynamic and efficient introduction
and removal of network resources as needed.

Resilience and The 6G E2E system requires pervasive service management and Trustworthiness
availability orchestration (e.g., high resilience and availability), encompassing
a comprehensive framework for data analysis, AI integration, and
the coordination of RAN functions, transport network functions,
5G/6G CNFs. E.g., separation of control plane (CP) and user plane
(UP), resilient mobility solutions, enhanced redundancy and
recovery mechanisms.

Persistent The objective is to establish a comprehensive framework in the 6G Trustworthiness

security and E2E system that ensures security and privacy are integrated across
privacy all components with the goal of assuring a trustworthy
environment. E.g., address current as well as future threats in a
resilient manner and incorporate security fundamentals in its
design, inherently support the preservation of privacy, allow

Dissemination level: Public | 15

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

different levels of anonymity for future services.

Internal This effort centres on the deployment of cloud-native virtual Economic

interfaces are network functions, emphasizing the development of exposure sustainability,
cloud interfaces that facilitate seamless communication between environmental
optimized different layers of the 6G E2E system. sustainability,
Trustworthiness

Separation of This refers to the optimized functionality in CN and RAN with Trustworthiness
concerns of bounded context and no duplication, avoiding complex
network interdependencies and cross-functional signalling- Self-sustained
functions NFs with minimal dependency on other network functions.

Network This initiative aims to avoid many standardized deployment Economic

simplification options and protocol splits. It also involves the evolution of the 5GC sustainability,
in comparison to accommodate the requirements of 6G RAN. The focus is on environmental
to previous simplifying protocols and minimizing User Equipment Network sustainability,
generations (UENW) signalling. Inclusiveness

Minimize This principle aims for E2E orchestration, emphasizing energy- Economic
environmental efficient and cost-conscious operations. It involves the sustainability,
footprint and implementation of a pervasive data and analysis framework environmental
enabling alongside the modularization of network functions. The sustainability,
sustainable infrastructure layer in the 6G E2E system should optimize both
use cases energy consumption and costs for enhanced sustainability and
operational efficiency of the use cases.

2.2 SYSTEM BLUEPRINT

Starting from the use cases, the underlying architecture design principles and the ecosystem
environment, the needed capabilities and requirements can be extracted and compared with a
system performance as well as in light of key values to show that the majority of the 6G use cases
have a significant positive impact on e.g., social, economic and environmental sustainability. This
will be an iterative design process wherein enablers tackling various aspects of the 6G system are
then analysed for their seamless integration into the 6G E2E system blueprint. This approach
aims to create a system that aligns closely with the real-world needs and expectations of users
and stakeholders, fostering a more adaptive and responsive design.

Consequently, the 6G E2E system blueprint, as discussed and further detailed in [HEX225-
D25], is depicted in Figure 2.1, wherein several novel aspects are encompassed.

Dissemination level: Public | 16

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 2.1: E2E System Architecture

The infrastructure layer encompasses all the E2E infrastructure and resources, physical or
virtualized, spanning across different technological domains and administrative domains. It
introduces the device-edge-cloud continuum, shorten as cloud continuum principle, which will
be essential for the 6G E2E system.

The network layer comprises beyond communications functions as well as the 6G RAN and
Core network functions blocks. Considering the migration principles, the 6G radio access should
support a single-RAT architecture only, i.e., a 6G UE that connects via the 6G radio interface
establishes a connection to the CN for 6G without any complex inter-RAT multi-connectivity, as
compared to the 5G paradigm. The 6G Core network function will be an extension of 5G Core
network instead of a subset. The 6G Beyond Communications Functions depicts the
functionalities to realize new services expanding beyond the communication capabilities.

The Network-centric application enablement platform layer and the 3rd party application
layer enable the full ecosystem applications to interact with the underlying network. The
application enablement platform layer Network-centric application layer is the main access to the
services of the 6G system as a platform which is controlled by the service exposure. Its main role
is to provide applications with easy access to 6G capabilities (including management services)
through abstracted APIs from the network and M&O services provide a further abstraction level
to simplify their use for non-experts and augmented with additional services from e.g., the
network applications, that can be leveraged by the developer of the applications of the
Application layer. , It is providing developer portal, operation, administration, and management

Dissemination level: Public | 17

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

(OAM) services (e.g., API discovery, ordering, monitoring). for API consumers, and common
functionalities for monetization, authorization, and legislation compliance. Network applications
can augment the 6G capabilities with additional services, provided also through APIs.

Aggregation of services, i.e., the action to compose services from a set of capabilities of
different 6G service providers, e.g., over multiple networks, is also provisioned in this layer.

The pervasive functionalities facilitate the four layers of the 6G system blueprint to realize the
full potential of the 6G platform, either jointly or independently. Both data management and AI
frameworks highlight the AI-centric approach to the 6G system as compared to the 5G system.
The data management framework integrating DataOps plays a crucial role in ensuring the
effective collection, processing, governance and quality of data, which is essential for pervasive
AI/ML operations. The 6G system architecture integrates AI/ML into every layer. The AI/ML
framework provide a comprehensive suite of tools and components designed to enable advanced
AI/ML workflows across the entire ecosystem. It consists of an AI orchestrator, AI/ML catalogue,
and MLOps orchestrator, working together to manage the lifecycle of AI/ML services, ensuring
efficient deployment, optimization, and continuous improvement. and integrate respectively
DataOps and MLOps. The management and orchestration framework moves beyond the
traditional relevant functions in 5G networks towards a more intent-based management
approach. Hence, the proposed blueprint introduces the Intent based Management framework as
one of the pillars of the 6G management and orchestration framework. The multi-platform
orchestration functionality provides a unified management and orchestration of network
services and network applications over a cloud continuum across multiple domains, owned and
administered by different stakeholders, and characterized by underlying heterogeneous
technologies platforms. The possibly AI/ML based closed loop controls are essential for an
increasing level of M&O automation toward autonomy in the 6G network operations. Network
Digital Twins (NDTs) have emerged as a tool for applying orchestration actions and observing
their effects before carrying them out on the real network. To provide accurate feedback to the
M&O, the NDT management creates the Network Digital Twin based on data collected from the
real network to accurately reflect the state and behaviour of the network [HEX225-D65]. A
broader security and privacy framework for the 6G E2E system is required as against a more
localized and domain specific approach on different layers of the 5G system [HEX225-D25].

Lastly, there will be multiples interactions between the various sets of stakeholders of the 6G
ecosystem. As the market is becoming more disaggregated, multiples stakeholders will be
engaged in the value creation of the 6G platform, moving away from a linear value chain toward
a multi-sided value chain. This leads to the definition of new roles as already described in
[HEX223-D22] and multiples multistakeholder interfaces between the various layers and

Dissemination level: Public | 18

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

domains of the 6G platform, .in particular, with federation APIs representing the functionalities
required for establishing and managing collaboration between different service or resource
stakeholders.

2.3 MIGRATION ASPECTS

In the migration from 4G to 5G, there was a big push for a gradual transition from 4G RAN
(LTE) towards a new 5G RAN, including a transition from 4G CN towards a new 5G CN (5GC). The
main reason was to allow the operators to flexibly upgrade RAN and CN independently. One
solution for this was to enable E-UTRA (i.e., 4G) and NR (i.e., 5G) Dual Connectivity, also known
as EN-DC. This created a plethora of EN-DC combinations, that each had to be standardized,
configured and tested. In the end, only a few combinations were eventually deployed, the main
option was when NR connected to 5GC and where LTE acted as Master Node (MN) and NR as
Secondary Node (SN) connected to EPC. The use of EN-DC as a migration and interworking
solution between 4G and 5G also delayed the introduction of 5GC since NR could be connected to
the EPC. This delayed several new 5G services related to the 5GC. Therefore, to avoid this complex
and also to enable a smoother and faster introduction of 6G services, the main option for
migration between 5G and 6G is to use a so called “evolved 5GC, E-5GC”, see Figure 2.1.

E-5GC allows the re-use of existing 5GC NFs, while introducing new dedicated 6G NFs to
support new 6G functionality. In addition, Multi-Radio Spectrum Sharing (MRSS) as a spectrum
migration solution can support interworking between 5G and 6G [HEX223-D43], see Figure 2.2.

Figure 2.2: Overview of the migration aspects

Dissemination level: Public | 19

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

For MRSS, the legacy 5G UEs should be able to avoid all 6G signals, i.e., the 6G signals are
“hidden” by reusing 5G locations in the time frequency grid. This means that the 5G UEs will not
be affected by these 6G signals. For example, the 6G CSI-RS locations should be a superset of 5G
CSI-RS time/frequency locations. Further on, the impact on the network performance and
updates to the 5G standard configuration should be minimized. From the 6G perspective, MRSS
should preferably not restrict design and operation of a 6G-only carrier and new 6G UEs should
be able to avoid 5G signals. It is expected that the overhead using spectrum sharing will be very
low and clearly lower than the corresponding spectrum sharing between 4G and 5G (DSS)
[HEX225-D35].

Taking the considerations of the migration aspects and using evolved 5G CN, the expected the
main interfaces between the UE, RAN and CN in 6G is depicted in Figure 2.2. As can be seen, a
Lower Layer Split (LLS) in the RAN will likely be used for 6G, instead of the High-Level Split (HLS)
used in 5G. The main reason for this is to simplify the RAN architecture, as well as to support D-
MIMO. Furthermore, it is expected that the N2 may be kept. The main reason for this is smoother
migration with 5G using same anchor point [HEX225-D35]. However, the N2 interface may need
to be evolved compared to 5G in order to handle the demands for a more cloud friendly
environment.

Another aspect to consider is that 6G aims to support new services and use cases. Although AI
and ISAC are the most visible examples of new 6G services that will put new requirements on the
E2E network, there may be others not yet anticipated. Building upon the Service Based
Architecture (SBA) of 5G, 6G can target streamlined NF design by collocating or refactoring 5G
NFs as well as new 6G NFs. This streamlined design can target different aspects of the network,
e.g. reduced signalling, increased flexibility or communication overhead, etc. Modular networks
also bring out challenges in the interactions between different modules and orchestration of the
modules and entities.

2.4 MODULAR ARCHITECTURE DESIGN

From the network performance perspective, 6G is envisioned to exceed 5G performance which
puts additional requirements to the CN design. In addition to the well-known KPIs, such as lower
latency or high throughput, the 6G network needs to ensure streamlined operations with
decreased number of standardized APIs, network functions and signalling. Therefore, 6G needs
to further enhance the deployment and operational flexibility by revisiting the 5G CN network
function (NF) design where needed. Therefore, it is crucial to find a balance between the
granularity of network functions and the number of interactions between system elements. This
balance enables flexible and modular addition, update, and replacement of NFs. While a high level

Dissemination level: Public | 20

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

of granularity in the control plane has its advantages and disadvantages, understanding the
various granularity options and their impact on end-to-end performance is essential to fully grasp
the potential of network modularization. The evaluation and analysis of these designs that
compass different levels of modular granularity is in line with the pros and cons analysis
presented in Table 2.2.

Table 2.2: Advantages and disadvantages of high granularity modules

Advantages Disadvantages

Efficient resource scaling through scalability Need for defined interfaces for cross-vendor
management deployments

Easier identification and replacement of faulty Increased complexity in integration and testing
modules

Faster development of independent modules Higher management overhead

Flexible module placement in distributed Increased signalling and data exchange between
deployments modules

Efficient reuse of modules in different More memory transactions for context data
implementations management

A module can encompass different network functions, like microservices, and there are
multiple approaches to constructing a network module. Different modularity studies are
conducted to investigate the impact of redesigning 5GS NFs while varying the granularity of these
NFs [HEX225-D35]. One study, called procedure-based decomposition, suggests redesigning the
5G Core NFs so that each NF incorporates the services and processing logic to execute a specific
5G procedures such as UE Registration or PDU Session Establishment, etc. Another study, called
Modular UPF, suggests modularizing the 5G UPF to smaller components that includes Uplink
packet processing logic, downlink packet processing logic, and optional on-demand UPF features
such as lawful intercept, etc. On the other hand, it is also possible to preserve highly granular
network modules and dynamically allocate them based on the network traffic and required
services. Finally, it is also possible to customize both the granularity and the composition of
different modules according to the steady state characteristics of the traffic demand and various
KPIs. Each one of these methods brings different advantages and challenges, therefore the
selection of method needs to be carefully selected.

The high degree of flexibility is achieved via the interactions between different NFs in 5GC. In
6G, taking advantage of the findings of 5G, the inter-NF or inter-module integration needs to be
streamlined, i.e., by optimizing where possible, removing unnecessary interactions, and

Dissemination level: Public | 21

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

preserving the already optimum ones. This requires not only optimizing the inter-module
interactions but also streamlining the interactions between different entities of the network as
the modules can be deployed there (e.g., RAN, CN, edge, etc.).

2.5 REFERENCES
[HEX223-D22] HEXA-X-II Deliverable D2.2 – Foundation of overall 6G system design and preliminary
evaluation results, https://hexa-x-ii.eu/wp-content/uploads/2024/01/Hexa-X-II_D2.2_FINAL.pdf,
December 2023.
[HEX225-D25] HEXA-X-II Deliverable D2.5 – Final overall 6G system design, to be published in April 2025
[HEX223-D43] HEXA-X-II Deliverable D4.3 – Early results of 6G Radio Key Enablers, https://hexa-x-
ii.eu/wp-content/uploads/2024/04/Hexa-X-II_D4_3_v1.0_final.pdf April 2024
[HEX225-D35] HEXA-X-II Deliverable D4.3 – Final architectural framework and analysis, https://hexa-x-
ii.eu/wp-content/uploads/2025/03/Hexa-X-II_D3.5_v1.0.pdf, February 2025
[HEX225-D65] HEXA-X-II Deliverable D4.3 – Final Design on 6G Smart Network Management Framework,
https://hexa-x-ii.eu/wp-content/uploads/2025/02/Hexa-X-II_D6-5_final.pdf February 2025

Dissemination level: Public | 22

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

3 EXTENSIONS
This section focuses on additional innovations and considerations needed to enhance the
foundational concepts of 6G architecture. These extensions address the integration of emerging
technologies, the scalability of existing frameworks, and the adaptation of network functionalities
to evolving user demands and environmental constraints. By building on core elements such as
resource optimization, security, and interoperability, these extensions aim to future-proof the 6G
ecosystem. Key focus areas include expanding the role of AI-driven network intelligence,
enhancing the modularity of network functions, and enabling seamless integration with non-
terrestrial networks and quantum computing paradigms. Together, these advancements position
the 6G architecture to not only meet but exceed the complex demands of next-generation
applications and services

3.1 INTEGRATION OF OVERARCHING CONCEPTS INTO

THE 6G ARCHITECTURE
This section explores the integration of foundational concepts that define the 6G architecture,
emphasizing the seamless connectivity and interoperability across heterogeneous networks. Key
aspects include the unification of distributed sub-networks into a cohesive "network of
networks," the adoption of advanced virtualization techniques, and the enhancement of resource
management through AI-driven frameworks. These concepts aim to address the growing
demands of ultra-reliable, low-latency communication, and energy efficiency while supporting
dynamic and adaptive network functionalities. By incorporating these overarching principles, the
6G architecture aspires to provide a robust, scalable, and future-ready framework capable of
meeting the diverse requirements of next-generation applications.

3.1.1 RESOURCE MANAGEMENT IN THE DEEP -EDGE-EDGE-CLOUD

CONTINUUM

The softwarization of entities at the deep edge (such as vehicles or robots), combined with
their increasing levels of automation and the virtualization of their components, is placing
growing flexibility and reliability demands on their networks. These networks must support a
rising number of compute-intense applications, functions and control processes (e.g. for
autonomous driving) which cannot be efficiently accomplished by only scaling and re-
dimensioning the networks at the deep-edge. The (cost-)efficient execution of these compute-
intense processes can be enhanced by leveraging the envisioned 6G deep-edge – edge – cloud
continuum, which enables the opportunistic offloading and distribution of processing tasks

Dissemination level: Public | 23

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

within the deep-edge and to the edge or cloud [6GSHINE24-D42]. Achieving this vision requires
seamless connectivity and integration of networks at the deep-edge with the 6G parent network.
This integration can be facilitated through the (gradual) adoption of wireless sub-networks in
entities such as vehicles or robots [reference to our previous section ‘Sub-networks integration
in the 6G Network of Networks’]. For instance, in the scenario of in-vehicle wireless sub-
networks, this vision would enable computing entities at the edge or cloud to function as a virtual
Electronic Control Unit (ECU), elastically extending the computing and processing capabilities of
the in-vehicle network and Electrical/Electronic (E/E) architecture using edge and cloud
resources [LCG+24].

APP 1 APP i APP n

Application to Network APIs

Network Intelligence Other core Network Communication &

functions (NIF) network Exposure Computational resources
Analytics / Predictions/… functions Function management functions (CCRMF)
Deep-edge – Edge – Cloud continuum
Deep-edge continuum
Computing & communication resources exposure functions (CCREF)

Edge

SN Deep-
edge 6G PN Cloud

Figure 3.1: System architecture for offloading with dynamic and coordinated resource
management in the 6G continuum.

Beyond the architectural components and their associated sub-network control functions
required for integrating sub-networks with the 6G parent network [LCG+24], current 5G and
beyond networks lack the components and interfaces necessary for a joint
orchestration/management and context-aware operation of infrastructure (i.e., computing &
communication links), network functions and application layers (see Figure 3.2) for the effective
and dynamic completion of compute-intense (control) process in the deep-edge – edge – cloud
continuum.

6G sub-networks at the deep edge will feature distributed communications and processing
capabilities for autonomous local data management. They are also being designed to seamlessly
integrate with the 6G parent network, forming a deep-edge – edge – cloud continuum. This

Dissemination level: Public | 24

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

integration facilitates an elastic continuum orchestration/management approach where

computing tasks and communication links are dynamically and jointly scheduled across the nodes
and links forming the continuum. The effective and efficient joint management of connectivity
and computing resources across the continuum necessitates of the following architectural
innovations (see Figure 3.2):

• Communication & Computing Resources Exposure Function (CCREF). 5G SA Network

Exposure Functions (NEF) mainly focus on the interface between Network Functions (NF)
and Application Function (AF) for traffic management and QoS assignments, through
interaction with the policy elements (i.e., Policy Control Function – PCF). In other words,
NEF exposes the application layer for interaction with the network function layer. CCREF
necessitates of open and harmonized interfaces for the exposure of resources and
capabilities of any type from the infrastructure layer across the 6G ‘network of networks’
(i.e., sub-network, 6G parent network, cloud), including computing and communication
resources, but potentially others like AI. CCREF is a clear architectural enabler for forming
the continuum which could be locally exploited at the deep-edge (deep-edge continuum)
and through the end-to-end (deep-edge – edge – cloud continuum). Network
functionalities implemented in the end-to-end system are also necessary to collect those
resources and capabilities and make them available to other network functions.

• Communication & Computational Resources Management Function (CCRMF). Extending

the capabilities of the 5G PCF, and through the interaction with CCREF, CCRMF enables
the implementation of advanced resource management policies that coordinate the
scheduling and allocation of the communication and computing resources through the
continuum (either at the deep-edge or end-to-end).

• Network Intelligence functions (NIF). Network Data Analytics Function (NWDAF) in 5G

networks leverages network data analytics to generate real-time operational intelligence
driving network automation and service orchestration. NIF complements NWDAF by
exploiting the CCREF (also NEF from application to network functions layers) and
deriving AI-driven proactive network management solutions in CCRMF, e.g., through the
prediction of the availability of computing and connectivity resources in the deep-edge –
edge – cloud continuum.

Dissemination level: Public | 25

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

R
R
G M
SNE W
O
R F
SNE R F
M

6G BS
6G PN
SN SNE

G
W
Control
R
G R Functions
W M
O Network Intelligence
F
F S functions+
N
M Analytics / Predictions/…
LC
Communication &
SNE Computational resources
G management functions
W SNE
Computing &
SNE communication resources
exposure functions
SNE SNE
SNE Network Exposure Function
SN
SN
EN

EN Entity SN Sub-network 6G PN 6G Parent network 6G Base Station 6G Core Network Edge/Cloud compute node
6G BS

High capabilities device Low capabilities device Subnetwork element Sub-network control function instances
LC SNE

S R O
N
M
Subnetwork management role G
W Gateway role R
M Radio Resource Management role F
F
Compute offloading role

Figure 3.2: Example of realization of the 6G architecture enablers identified in Figure 3.1 for
offloading with dynamic and coordinated resource management in the 6G continuum.

Figure 3.2 show an example of the realization of the system architecture depicted in Figure 3.1
using the sub-network integration with 6G parent network Figure 3.1. For guaranteeing the
survivability of sub-networks when the integration with the 6G parent network cannot be
established, network functions can flexibly be deployed/instantiated in the sub-network nodes
depending on their capabilities. For the isolated operation of sub-networks, Higher Capabilities
(HC) devices with SNM/OFF/GW roles are entitled to jointly manage the communication and
computing resources exposes by other elements of the sub-network (e.g., LC) to the CCREF
following the policies and procedures of the CCRMF and satisfying the service requirements of
the SNEs exposed through the NEF. The flexible deployment of the network function layer
functions also allows for the centralized and partially decentralized (via functional splitting
between the sub-network and the 6G-parent network) management of computing and
connectivity resources of the deep-edge – edge – cloud continuum.

3.1.2 SUSTAINABILITY AND COMPUTE CONTINUUM

The interaction of network functions with the underlying infrastructure may pose significant
inefficiencies if not properly performed [ORIG24-D21].

The virtualization of Radio Access Networks (vRANs) offers numerous benefits, such as
reduced vendor lock-in and resource efficiency, yet it faces significant challenges, particularly
with latency-sensitive tasks like LDPC decoding. Current solutions often rely on expensive and

Dissemination level: Public | 26

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

energy-intensive hardware accelerators, such as ASICs and GPUs, to meet the strict performance
requirements of 5G, which raises concerns about sustainability.

Additionally, the integration of 6G virtual network functions (VNFs) with O-RAN's Radio
Intelligent Controller (RIC) is complicated by poor interoperability among network components,
leading to potential conflicts and inefficiencies. The disaggregated nature of O-RAN, coupled with
the need for real-time AI-driven decision-making, further exacerbates these challenges.

Figure 3.3: The compute continuum layer

Network Intelligence (NI) solutions in 6G also struggle with latency and inefficiency, often due
to a misalignment between AI/ML algorithms and the underlying network infrastructure. This
misalignment can result in excessive data transmission and suboptimal performance. A more
tailored approach that leverages edge computing and optimizes the interaction between NI
algorithms and network infrastructure is necessary for improving efficiency and responsiveness.

Finally, the potential of modern programmable transport technologies, like smartNICs and
Network Processing Units (NPUs), is underutilized due to the complexity of programming in-band
computing models. These limitations prevent the full exploitation of these technologies,
hindering the development of innovative 6G services that require real-time processing at line
rates. Simplifying the deployment of sophisticated user-plane VNFs is crucial for unlocking new
applications and markets.

The Compute Continuum Layer (CCL) [ORIG24-D21] depicted in Figure 3.3 is an architectural
innovation designed for 6G systems, facilitating the execution of network processing workloads
across diverse computing resources. This architecture aims to streamline resource sharing and
fully utilize the capabilities of a heterogeneous computing environment, encompassing GPUs,
TPUs, FPGAs, ASICs, NPUs, smartNICs, and quantum computers, alongside traditional CPUs.

Dissemination level: Public | 27

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

The CCL introduces compute-aware network operations while preserving the abstraction of a
pure virtualization layer. This allows network functions (NFs) to be optimized by matching them
with appropriate computing resources, leading to enhanced efficiency and reduced resource
usage. Real-time resource management is a key feature, enabling dynamic allocation and policy
enforcement to maximize resource utilization across the edge-to-cloud continuum.

A centralized abstraction layer within the CCL simplifies the interaction between NFs and
hardware components, improving resource management and scalability. Additionally, the CCL
accommodates new network sensing functionalities, benefiting tenants interested in metadata
over data transport. By regulating virtualized network functions and supporting operational
decisions, the CCL ensures their correct performance, particularly in scenarios with stringent
time constraints, such as virtualized radio access networks (vRANs).

The CCL provides a robust API that abstracts the heterogeneous and disaggregated computing
infrastructure, making it accessible to the 6G network's software components. This API also offers
tools for exploiting the underlying infrastructure and implementing policies that govern its usage.
This approach prevents vendor lock-in while ensuring optimal performance in the demanding 6G
environment.

The CCL architecture addresses critical challenges in 6G networks, such as unsustainable RAN
virtualization, poor interoperability, high latency in processing complex network tasks,
underutilized programmable transport, and inadequate data representation. It achieves this by
accessing various physical processing units through an abstraction layer, which homogenizes
these resources as Logical Processing Units (LPUs). During operation, the CCL manages
processing requests from NFs, ensuring that tasks are efficiently routed and executed within set
policies and deadlines, thereby optimizing cost and energy consumption while maintaining
required performance levels.

3.1.3 ZERO-TRUST SECURITY AND INTEROPERABILITY

The current telecommunications landscape faces challenges due to the lack of standardized
global service APIs and the siloed development of network management, orchestration, and
control functions. This fragmentation hinders cross-domain interaction, limiting the effectiveness
of network slicing, especially in business-to-business contexts. As network slicing becomes more
crucial, the need for modular, interoperable network functions (NFs) and flexible data exchange
across domains grows. The adoption of a publish-subscribe methodology and enhanced
configurability of NFs are essential steps toward achieving real-time, automated optimization
using AI and big data solutions [ORIG24-D21].

Dissemination level: Public | 28

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Global mobility for IoT devices also presents challenges, as current international roaming
models rely on outdated, trust-based agreements between mobile network operators (MNOs),
leading to performance penalties and increased costs. A decentralized identity model that
decouples user authentication from home operators could enable more efficient local breakout,
reducing the need for costly international data routing.

Moreover, the current approach to privacy, security, and data representation in cellular
networks is insufficient for the demands of global connectivity and IoT deployments. Improved
data governance and high-quality data provisioning are critical for supporting advanced network
intelligence (NI) functionalities and enabling precise management of infrastructure.

Finally, while the service-based architecture (SBA) introduced in 5G offers flexibility, it has led
to a surge in signalling traffic, challenging scalability and cost efficiency. Optimizing signalling
management is vital as networks evolve to support an increasing number of connected devices
and complex IT integrations.

To meet the demands of future global connectivity, there is a need for a comprehensive
overhaul of network architectures, focusing on standardized APIs, modular NFs, decentralized
identity models, and efficient signalling management. These improvements are essential for
unlocking the full potential of 6G networks. Simplifying the deployment of sophisticated user-
plane VNFs is crucial for unlocking new applications and markets.

The Zero-Trust Layer (ZTL) [ORIG24-D21] aligns service providers' internal operations with
network operators' continuous optimization efforts, thereby unlocking advanced functionalities
like remote sensing and digital twinning. By fostering a cooperative control loop, the ZTL enables
service providers to have a more direct influence on network operations, while ensuring privacy
and security—like how hyperscale cloud services operate today.

The ZTL offers both vertical and horizontal exposure. Vertically, it enhances network analytics
by integrating feedback from service providers into the Network Data Analytics Function
(NWDAF), enabling more precise customization without compromising confidential information.
This approach allows service providers to optimize their own metrics in line with network quality
of experience (QoE), which may differ from standard network metrics due to business-specific
factors.

Horizontally, the ZTL supports global operations, particularly for IoT devices, by facilitating
efficient international roaming and enabling new business models. The architecture envisions a
decentralized identity system, decoupling user authentication from the connectivity services
provided by home operators. This allows visited operators to directly charge global end-users
while giving home operators full visibility into these transactions. The ZTL also incorporates

Dissemination level: Public | 29

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

distributed ledger technology for secure, immutable record-keeping and network intelligence
modules for real-time anomaly detection, enhancing the security, privacy, and operational
efficiency of global networks. This forward-thinking architecture aims to transform traditional
network interactions and meet the high demands of next generation 6G networks, ensuring
seamless, global connectivity for a diverse range of devices and services.

Figure 3.4: The Integration of the ZTL in the Overall 6G Network Architecture.

3.1.4 SUB-NETWORK INTEGRATION IN 6G NETWORKS

The softwarization of entities at the deep edge (such as vehicles or robots), combined with
their increasing levels of automation and the virtualization of their components, is placing
growing flexibility and reliability demands on their networks. These networks must support a
rising number of compute-intense applications, functions and control processes (e.g. for
autonomous driving) which cannot be efficiently accomplished by only scaling and re-
dimensioning the networks at the deep-edge. The (cost-)efficient execution of these compute-
intense processes can be enhanced by leveraging the envisioned 6G deep-edge – edge – cloud
continuum, which enables the opportunistic offloading and distribution of processing tasks
within the deep-edge and to the edge or cloud [6GSHINE24-D42]. Achieving this vision requires
seamless connectivity and integration of networks at the deep-edge with the 6G parent network.
This integration can be facilitated through the (gradual) adoption of wireless sub-networks in
entities such as vehicles or robots, as discussed in Section 3.1.1. For instance, in the scenario of
in-vehicle wireless sub-networks, this vision would enable computing entities at the edge or
cloud to function as a virtual Electronic Control Unit (ECU), elastically extending the computing

Dissemination level: Public | 30

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

and processing capabilities of the in-vehicle network and Electrical/Electronic (E/E) architecture
using edge and cloud resources [LCG+24]. Beyond the architectural components and their
associated sub-network control functions required for integrating sub-networks with the 6G
parent network [LCG+24], current 5G and beyond networks lack the components and interfaces
necessary for a joint orchestration/management and context-aware operation of infrastructure
(i.e., computing & communication links), network functions and application layers (see Figure
3.1) for the effective and dynamic completion of compute-intense (control) process in the deep-
edge – edge – cloud continuum.

3.2 INTEGRATION OF 6G NETWORK PARADIGMS

This section discusses the convergence of 6G network paradigms, focusing on the seamless
blending of network slicing, multi-access edge computing (MEC), and cloud-edge continuum
strategies. The integration of these paradigms ensures enhanced performance, resource
efficiency, and service customization to meet the heterogeneous demands of advanced
applications. By aligning these paradigms under a unified framework, 6G networks can provide
dynamic, scalable, and intelligent services, supporting innovative use cases such as real-time
extended reality, autonomous systems, and smart infrastructure. This integration is critical to
unlocking the full potential of 6G technologies and achieving global connectivity goals

3.2.1 NETWORK SLICING AND MULTI-ACCESS EDGE COMPUTING

(MEC)

Network slicing is a key feature of 5G and beyond systems to simultaneously support multiple
services with heterogeneous requirements (e.g., data rates, latency, reliability, availability). It
allows sharing a common infrastructure by creating multiple end-to-end logical networks,
denoted as network slices, on top of it. Each network slice is optimized in accordance with the
requirements of a set of services.

While the network slice concept is already quite consolidated, a new dimension of slicing that
encompasses the edge computing domain has been recently introduced by the ETSI MEC
standardization group with the so-called MEC Application slices [ETSI22-MEC038], [ZLL+22]. This
concept arises from a customer-oriented perspective, where the customer uses a virtualized
application (e.g., XR application, AI application), referred to as MEC application (MEC App), that
needs to run at the MEC system to perform computations. In this case, the MEC App cannot be
only considered as a part of the network slice, since its requirements go beyond the network
requirements (i.e., data rate, latency, reliability) to include other requirements such as computing
resources, isolation at application level, virtualization approach (e.g., deployment as virtual

Dissemination level: Public | 31

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

machines, containers), etc. This motivated the introduction of the new MEC application slice (MEC
App slice) as an independent entity from the network slices that relies on similar concepts
regarding isolation and QoS guarantees but adapted to the MEC system [ZLL+22].

Regarding the network slice management, the 3GPP has defined a general architectural
framework for management and orchestration [3GPP24-28.533] [3GPP23-28.530], while the
concepts of network slice management are defined in [3GPP23-28.530]. Although the 3GPP
architecture for network slice management in [3GPP24-28.533] is general and gives room for
different implementations, a commonly considered approach is given in [BTB22] and is taken as
a reference for the proposed slice manager in Figure 3.5. Specifically, the management of network
slices is based on three functions:

• Communication Service Management Function (CSMF): This function is the user interface
for slice management and converts the Service Level Agreements (SLAs) (i.e. the business
contract between the service provider and the client that specifies the service levels to be
ensured) into the Service Level Specification (SLS), which includes the set of technical
attributes that have to be satisfied by the network slice.

• Network Slice Management Function (NSMF): This function is responsible for the
management and orchestration of the NSI to fulfil the SLS specified by the CSMF. This
includes the different stages of the lifecycle management of an NSI, namely the
commissioning (i.e. the slice creation, in which the necessary resources are allocated and
configured based on the SLS requirements), the operation (i.e. the activation, supervision,
performance reporting, modification, and de-activation of the NSI) and the
decommissioning (i.e. the termination of a NSI when no longer needed).

• Network Slice Subnet Management Function (NSSMF): The NSMF splits a NSI into its
subnet slice instances, i.e. RAN slice, TN slice and CN slice, indicating for each one the SLS
to be fulfilled. Then, there is a NSSMF taking care of the lifecycle management of each
subnet slice, namely the RAN NSSMF, the TN NSSMF and the CN NSSMF depicted in Figure
3.5.

Regarding the MEC App slice management, the following two functionalities defined by ETSI
MEC in [ETSI24-MEC044] are considered:

• MEC Application Slice Communication Service Management Function (MAS-CSMF): This

function is responsible for translating high level service-related QoS requirements into
MEC App slice requirements. Moreover, it also facilitates the purchase and monitoring of
MEC App slices from the customer side, e.g. through the exposure of service performance
and alarm information.

Dissemination level: Public | 32

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

• MEC Application Slice Management Function (MAS-MF): This function takes care of the
design of the MEC Application Slice template (MAST) and of the lifecycle management of
the MEC App slice instances according to the requirements specified by the MAS-CSMF.
The MAST is a collection of parameters that define an information model including the
MEC App slice identifier, the name, the designer, the version, the release time and the
description. In turn, the lifecycle management consists in the creation, activation,
operation and release of the MEC App slice instances across the MEC system. This can
involve aspects such as the selection of the most appropriate MEC hosts at the edge sites
of the edge-to-cloud continuum to fulfil the QoS requirements established in the MAST.

Slice manager
Network slice management MEC App slice management

CSMF MAS-CSMF

NSMF MAS-MF

RAN TN CN
NSSMF NSSMF NSSMF

Figure 3.5: Network Slices Management in the Edge

This joint vision of MEC App slices and network slices is illustrated in Figure 3.6, which shows
an example with two services, namely an XR service and holographic communications service.
Each one is supported by a different network slice, composed by RAN, transport network (TN)
and core network (CN) subnet slices, and by a MEC App slice available at the MEC host of an edge-
site, which in this case is co-located with the gNB. Each MEC App slice contains the required MEC
Apps to support the task computation associated to the service. The gNB can forward tasks to the
MEC host through the local breakout mechanism, which selects the IP-based traffic to be
forwarded to the local User Plane Function (UPF) and from then to the MEC host. For instance,
the rendering tasks of a user of the XR service are sent to the gNB through RAN slice #i, and then
forwarded to the MEC host, which processes them in the rendering MEC App of MEC App slice #X.

As seen in Figure 3.6, the slice manager needs to interact with other architectural components
of the orchestration, management and control layer in order to support the operation of the
different management functionalities. Specifically, the operation of the RAN slices conducted by
the RAN NSSMF involves the allocation of the radio resources at the different cells to the RAN
slices. For this purpose, the RAN NSSMF can rely on the RAN intelligent controllers to conduct
this allocation based on specific algorithmic solutions that can involve the use of AI models.
Depending on the solution, this can be done with RAN controllers having a global scope involving

Dissemination level: Public | 33

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

multiple gNBs, or having a local scope for solutions involving the allocation of resources to slices
separately to different gNBs.

Figure 3.6: Joint vision of MEC App slices and network slices

For the MEC App slice management, the MAS-MF of the slice manager needs to interwork with
the MEC orchestrator (MEO) functionality considered by ETSI MEC in [ETSI24-MEC044]. This
orchestrator has an overall view of the MEC system across the different edge sites of the edge-
cloud continuum (i.e., deployed MEC hosts, available resources, MEC services and topology) and
will take care of supporting the MEC App slice lifecycle management operations of the MAS-MF,
e.g. through onboarding application packages and selecting appropriate MEC hosts for the
instantiation of MEC App instances.

Dissemination level: Public | 34

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

3.2.2 TIME-CRITICAL AND DETERMINISTIC COMMUNICATIONS

End Det. Devices/ 6G Access 6G Core Det. Data Edge

devices Local gateways Network computing
NW.
Applications
1
Application Middleware
2 End-to-end Deterministic Networking 6 2 (TSN / DetNet)
6
Management, Orchestration, Monetization
5
6
Access, Mobility, Network applications
3 5
Cloud infrastructure, Data pipeline
4

Transport

6G System

Figure 3.7: Integration of DetNet components in the network architecture.

Ensuring dependable end-to-end time-critical communications across diverse communication

and compute domains is critical for supporting emerging applications, such as extended reality
(XR), smart farming, and adaptive manufacturing [DET6G23-D11]. These applications demand
stringent performance metrics and guarantees, particularly in terms of hyper-reliable and low-
latency communications. However, despite advances in latency and reliability, both computation
(e.g., edge computing) and communication (e.g., 5G URLLC) domains exhibit substantial
stochastic variations compared to wired deterministic communications technologies like TSN
and DetNet, particularly in terms of delay characteristics and non-negligible packet delay
variations. This leads to the traditional approach for end-to-end resource planning falling short
regarding service performance, scalability and efficiency, especially when integrating stochastic
elements and deterministic communication technologies to support time-critical services
[DET6G23-D22] [DET6G23-D31]. Future networks must address these challenges fundamentally
in the design, planning, and operation of time-critical networks, focusing not on achieving purely
deterministic performance at the cost of resources but on embracing stochasticity while ensuring
application dependability.

In reference to a horizontal 6G architecture, the following areas need attention (as indicated
in Figure 3.7 above and described in [DET6G24-D12]):

1. Support for end-to-end time-critical applications [DET6G23-D11].

2. Advanced network configuration via network exposure, to invoke dependable

communication in accordance with a well-defined service specification for requesting a
dependable communication service from the network [DET6G24-D12].

Dissemination level: Public | 35

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

3. The 6G network needs to provide a dependable communication service, which means that
it must be able to comply with and deliver the performance that is requested from the
applications. This includes being able to monitor the KPIs that characterize the delivered
service performance and provide a basis for data-driven (latency) performance
prediction, so that the 6G network can determine which (latency) performance levels it
can promise to what reliability level [DET6G23-D21]. One important characteristic is to
be able to control also the packet delay variation as explained in [DET6G23-D31].

4. Dependable time-critical communication builds on time-awareness throughout the

system, which is based on robust time synchronization [DET6G23-D22].

5. To enable edge computing-based realizations of time-critical services – including the

integration with deterministic TSN/DetNet networking – dependable time-aware
compute features need to be provided by the edge [DET6G24-D12].

6. When considering latency variations of sub-components in an end-to-end system, the

(TSN/DetNet) end-to-end traffic handling should be aware of the latency characteristics
of sub-components, in order to enable robust and optimized end-to-end deterministic
network configurations [DET6G23-D31].

The proposed approach for end-to-end dependable time-critical communication advocates (I)
the acceptance and integration of stochastic elements, such as wireless links and computational
elements, with a focus on characterizing their stochastic behaviour. Monitoring and predicting
KPIs, such as latency or reliability, which can be leveraged to make individual elements plannable
despite the presence of stochastic influences. Nevertheless, system enhancements to mitigate
stochastic variances in communication and computational elements are required, by means of
mechanisms such as packet delay corrections [DET6G23-D21]. (II) Furthermore, the entire end-
to-end interaction loop (e.g., the control loop from the sensor to the controller to the actuator) is
managed, considering the underlying stochastic characteristics, especially with the integration of
compute elements. (III) Due to unavoidable stochastic degradations of individual elements, the
approach considers adaptation between applications running on such converged network
infrastructures. In other words, application requirements can be adjusted based on prevailing
system conditions leading to more flexibility in the operations.

The approach builds on the concept of time-awareness by ensuring accurate and reliable time
synchronicity while also incorporating security-by-design for dependable time-critical
communications. Generally, the notion of deterministic communication, where the behaviour of
network, compute nodes, and applications is pre-determined, is extended towards dependable
time-critical communication, where the focus is on managing communication (and compute)

Dissemination level: Public | 36

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

characteristics to provide the KPIs and reliability levels required by the application. Architectures
and algorithms are facilitated for scalable and converged future network infrastructures that
enable dependable time-critical communication end-to-end, across domains, including 6G.

3.3 TOWARDS A GLOBAL SBA

This section outlines the vision for a Global Service-Based Architecture (GSBA) that unifies
service declaration and management across diverse domains, including network operators,
infrastructure providers, and application ecosystems. The GSBA aims to overcome trust and
interoperability challenges by leveraging modular, API-driven frameworks that enhance
scalability, flexibility, and automation. By integrating advanced conflict resolution mechanisms,
such as those used in RAN Intelligent Controllers (RIC), and ensuring seamless interaction
between heterogeneous network components, the GSBA paves the way for innovative business
models and efficient resource utilization. This architecture forms the backbone for enabling
global connectivity, unlocking new opportunities for collaboration and service delivery in the 6G
era.

3.3.1 GLOBAL SERVICE BASED ARCHITECTURE

Both the Cloud Continuum and the Zero Trust Functionality, along with other legacy domain
buses such as 3GPP SBA, rely on the Global SBA (GSBA). The GSBA is designed to facilitate the
declaration and management of services across different domains, such as network operators and
infrastructure providers. In this context, a "domain" corresponds to an entity that manages
various sub-domains, including the radio access network, core network, and international carrier
network.

A significant challenge within this ecosystem is the inherent lack of trust between these
entities, which hampers resource sharing and the adoption of innovative business models. While
legacy domain buses like 3GPP SBA provide essential support to GSBA, there are domains where
the development of new buses is necessary.

3.3.2 BRINGING THE GSBA TO THE RAN

One possible architectural option is to bring the Global SBA to the RAN. The RAN bus operates
within the RAN Intelligent Controller (RIC) platform, playing a key role in enabling the collection
of key performance measurements (KPM) from RAN nodes and facilitating RAN control (RC)
decisions for infrastructure management. The RIC platform hosts multiple xApps that utilize the
RAN bus for a variety of functions.

Dissemination level: Public | 37

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

For instance, xApps collect RAN node performance data using E2 service model KPMs (E2SM-
KPM), a service defined by the O-RAN Alliance. Additionally, xApps can adjust Information
Elements (IEs) within specific signalling messages through the RAN node using E2SM RAN
Control (E2SM-RC), avoiding the need to decode entire network messages.

Conflicts may arise within the RIC when xApps—especially those developed by third parties—
modify IEs in ways that are incompatible with each other. These conflicts are categorized as
direct, indirect, and implicit, each requiring specific resolution strategies.

The RIC’s conflict management mechanisms are critical in addressing these issues. Techniques
such as post-action verification and tailored approaches for managing indirect and implicit
conflicts ensure the smooth interoperability of xApps. By effectively resolving these conflicts,
third-party xApp developers can enhance RAN node performance, with each xApp focusing on
optimizing specific performance metrics.

Introducing SBA into the RAN brings several benefits to network operations and management.

• Improved Scalability: Service-based architectures (SBA) decouple network functions into

modular services, allowing networks to scale dynamically based on demand. This
flexibility is essential for handling traffic spikes or expanding capacity without significant
hardware investments.

• Enhanced Flexibility and Modularity: The modular design of SBA enables independent
development, deployment, and management of network functions. This approach
supports agile updates and innovation without disrupting the entire network.

• Improved Automation and Orchestration: SBA supports advanced automation tools and
orchestration frameworks. By using programmable interfaces and machine-readable
APIs, networks can automate tasks like resource allocation, fault detection, and recovery.

• Better Resource Management: Fine-grained control over individual services enables

better monitoring and allocation of resources.

• Resilience and Reliability: SBA supports fault-tolerant designs where failures in one
service do not cascade across the network. This architecture improves the overall
reliability and uptime of the network.

• Future-Proofing: With its modular and API-driven approach, SBA is well-suited to adapt
to evolving standards, protocols, and technologies, ensuring long-term relevance and
reduced need for overhaul

Dissemination level: Public | 38

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

3.4 REFERENCES
[3GPP23-28.530] 3GPP TS 28.530 v18.0.0, "Management and orchestration; Concepts, use cases and
requirements (Release 18)", December, 2023.
[3GPP24-28.533] 3GPP TS 28.533 v18.2.0, "Management and orchestration; Architecture framework
(Release 18)", June, 2024.
[6GSHINE24-D42] 6G-SHINE, Deliverable D4.2, - “Preliminary results on the management of traffic,
computational and spectrum resources among subnetworks in the same entity, and between subnetworks
and 6G network”, https://6gshine.eu/wp-content/uploads/2024/11/D4.2_Preliminary-results-on-the-
management-of-traffic-v1.0.pdf June 2024.
[BTB22] S. Bolettieri, D. Thai, R. Bruno, “Towards end-to-end application slicing in Multi-access Edge
Computing systems: Architecture discussion and proof-of-concept”, in Future Generation Computer
Systems, vol. 136, November 2022, pp. 110-127.
[DET6G23-D11] DETERMINISTIC6G, Deliverable 1.1, “DETERMINISTIC6G use cases and architecture
principles,” Jun. 2023, https://deterministic6g.eu/index.php/library-m/deliverables
[DET6G23-D21] DETERMINISTIC6G, Deliverable 2.1, “First report on 6G centric enablers”, Dec. 2023,
https://deterministic6g.eu/index.php/library-m/deliverables
[DET6G23-D22] DETERMINISTIC6G, Deliverable 2.2, “First Report on the time synchronization for E2E
time awareness,” Dec. 2023, https://deterministic6g.eu/index.php/library-m/deliverables
[DET6G23-D31] DETERMINISTIC6G, Deliverable 3.1, “Report on 6G convergence enablers towards
deterministic communication standards,” Dec. 2023, https://deterministic6g.eu/index.php/library-
m/deliverables
[DET6G24-D12] DETERMINISTIC6G, Deliverable 1.2, “First report on DETERMINISTIC6G architecture,”
April 2024, https://deterministic6g.eu/index.php/library-m/deliverables
[ETSI22-MEC038] ETSI GR MEC 038 v3.1.1, "Multi-access Edge Computing (MEC); MEC in Park Enterprises
deployment scenario", November, 2022.
[ETSI24-MEC044] ETSI GR MEC 044 v3.1.1, "Multi-access Edge Computing (MEC); Study on MEC
Application Slices", April, 2024.
[LCG+24] L. Lusvarghi, B. Coll-Perales, J. Gozalvez, K. Aghababaiyan, M. Almela, M. Sepulcre,
"Characterization of In-Vehicle Network Sensor Data Traffic in Autonomous Vehicles", Proceedings of the
15th IEEE Vehicular Networking Conference (VNC 2024), 29-31 May, 2024, Kobe, Japan.
[ORIG24-D21] ORIGAMI Deliverable D2.1, M. Skarp, “Initial report on requirements and definition of KVIs
and KPIs”, Zenodo, Jun. 2024. doi: 10.5281/zenodo.12580929
[ZLL+22] H. Zhu, J. Liu, Y. Lin, Q. Wang, "MEC Application Slice and Its collaboration with 5G network slice,"
2022 IEEE International Symposium on Broadband Multimedia Systems and Broadcasting (BMSB), Bilbao,
Spain, 2022, pp. 1-6.

Dissemination level: Public | 39

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

4 UBIQUITOUS NETWORKS
The 6G, aiming at widest possible access to digital services across the world, will provide near-
universal, seamless wireless connectivity across a wide range of locations, including both urban
and rural areas, essentially offering internet access almost anywhere on Earth, thanks to
technologies such as distributed (and cell-free) multiple-input-multiple-output (MIMO)
technologies for dense urban areas and the integration of technologies non-terrestrial networks
(NTN)for areas with pour terrestrial coverage such as rural. This will effectively eliminate
coverage gaps and will deliver consistent high-speed connections regardless of the user's
location.

Hence, ubiquitous connectivity is now an explicit challenge to be addressed by the IMT-

2030/6G system to which 6G technologies can contribute. In addition, resiliency is another
challenge to ensure maximum service availability where NTN can contribute to the overall 6G
system as back up network infrastructure to the terrestrial network component, however with
limited capacity for example to convey most critical traffic. The NTN component itself can be
designed for intrinsic resiliency through multi-layer infrastructure combining multi-orbit
satellite access (e.g. GSO, NGSO).

Fixed wireless access (FWA), and the integration of transport network with distributed MIMO
(dMIMO) and Cell-Free MIMO (CF-MIMO) systems is an innovative architecture to support high-
capacity coverage in the built-up urban environments. The use of sensing data, e.g., in Integrated
Sensing and Communications (ISAC) systems can improve the efficiency as well as extending the
network coverage.

4.1 UBIQUITOUS COVERAGE VIA 6G NTN

ARCHITECTURE
Compared to 5G-NTN, 6G-NTN can address the same use cases with higher service
performances and Quality of Experience (QoE). The QoE includes the terminal design that should
be adapted to the operational constraints.

The underpinning concept of 6G NTN is a 3D multi-layered architecture [6GNTN24-D35]. The

“3D” characteristic stems from the native unification at radio interface and access layers of the
non-terrestrial component with the terrestrial one, while the “multi-layered” feature is related to
the integration of different layers consisting of communication nodes, i.e., satellites or HAPs flying
at different altitudes.

Dissemination level: Public | 40

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Two types of non-terrestrial nodes could be considered, namely deterministic nodes with a
fixed and predictable orbit (both Geostationary and non-GSO in low Eart orbit) and flexible nodes,
namely HAPS or special heavy drones, which might or might not be present at different points in
time and at different locations to extend coverage or enhance the network capacity. The latter are
supposed to be deployed “opportunistically” depending on specific needs and they are not meant
to be a permanent infrastructure with global coverage. Target frequency bands are C-band for
low data rate services to UE with hemispherical antennas and Q/V-band for high data rate
services to UE with directive flat panel antennas. Moreover, optical Inter-Node Links (INL) are
foreseen, with the only exception of LEO-GEO links where a dedicated allocation in Ka-Band can
be exploited. It is envisaged to implement all RAN functionalities and eventually some core
network functionalities in space in order to reduce latency (especially on the control plane) and
allow also direct connectivity between UEs via satellite. To cope with mass and power restrictions
especially in the LEO segment, a distributed approach might be adopted, splitting RAN
functionalities across several satellites. Last not least, the 6G-NTN space segment shall provide
reliable UE positioning service.

Figure 4.1: The 6G-NTN architecture concept

The following applications can represent use cases for a unified 3D network integrating
terrestrial, airborne and satellite access layers:

• Flexible payload-enabled service provisioning to semantics-aware and delay-tolerant IoT

applications supported with low-density LEO satellites’ constellations with periods of
invisibility (both for feeder link and access link to the UE),

Dissemination level: Public | 41

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

• Direct connectivity to Smartphones: At stake is to enable a certain level of service in light

indoor environment and increase bandwidth service towards broadband, this could be,
e.g., unified RAN for direct handheld device access in the Ka band,

• Broadband connectivity to vehicle/drone mounted devices: At stake is to be able to serve

ultra-small aperture terminals adapted to the installation constraints on the cars and
drones support for air-space safety critical operations [6GNTN24-D35].

The design of the NTN component shall consider the following constraints

• Spectrum scarcity requiring improved coexistence between NTN and TN

• Sustainability considerations aiming at minimizing the NTN intrinsic footprint, enabling

NTN to minimize the 6G overall footprint and identifying how NTN can maximize the 6G
overall handprint (addressing some of the SDGs)

In addition to ensuring support for the highly dynamic topology and configuration variability
of the three-layer access network and effective management of intra- (horizontal) and inter-layer
(vertical) handovers challenged by this variability, several enablers must be provided:

• Unified MANO for E2E management of both terrestrial/non-terrestrial infrastructure (for

the latter, also using the flexible payload feature), the topology of mobile network
implemented in software, and services provided by this network;

• Mechanisms of AI to support E2E network performance optimisation, its predictive

analytics as well as horizontal and vertical handovers;

• Distribution of core network between terrestrial and non-terrestrial layers;

• Store and Forward mechanisms for support of delay-tolerant IoT applications served by
periodically intermittent satellite connectivity;

• E2E control of UPF based on multi-domain SDN;

• Semantics-aware analytics;

• Direct handheld access for UEs served both by terrestrial and non-terrestrial RAN,
without the need of external antennas application.

4.1.1 DISTRIBUTION OF CORE NETWORK BETWEEN TERRESTRIAL AND

NON-TERRESTRIAL LAYERS

Recent IoT advancements have created new markets and use cases, but these services are
mainly limited to urban areas with existing cellular coverage, leaving 85% of the Earth’s rural and
offshore regions disconnected. This lack of coverage restricts technological and social progress,

Dissemination level: Public | 42

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

limits business opportunities, and hinders the potential of massive IoT applications. Achieving
global connectivity is crucial for fully leveraging IoT technologies, and satellite-based non-
terrestrial networks (NTNs) are key in addressing this. While current satellite proprietary
solutions are not cost-effective, 3GPP has standardized cellular NTNs using terrestrial
technologies such as NR, eMTC, and NB-IoT. In Release 17, 3GPP focused on transparent payload
architectures, which require complex and costly mega-constellations [STK+24]. Low-density LEO
satellite constellations offer a simpler approach and can leverage regenerative payload for delay-
tolerant IoT applications, reducing costs and enhancing interoperability. The low-density LEO
satellite constellation provides large revisit times making it possible to have several messages
per day, which is sufficient for many delay-tolerant applications such as agriculture, livestock
monitoring, assets tracking and maritime. However, it presents challenges, as using the low-
density satellite constellation introduces service link discontinuities. Moreover, the feeder link
connecting the low-density satellite constellation and Ground Station is accessible only at a
limited number of locations. To address this, a regenerative payload hosting part of the core
network and using a Store and Forward (S&F) mechanism is proposed, necessitating adaptations
in 3GPP standards [KCC+22]. S&F mechanism is indispensable for both User Plane and Control
Plane and will enable the flexible payload to provide NB-IoT coverage to delay-tolerant
applications. Implementing these modifications requires adapting the 3GPP standard procedures.
Enabling NB-IoT services from low-density satellite constellations requires sustaining service
links without constant Ground Station connectivity, necessitating discontinuous backhauling.
Cost-efficient deployment also relies on standard 3GPP interfaces, allowing multiple service
providers to share LEO constellations and extend coverage using roaming agreements. A
distributed 3GPP architecture, focusing on regenerative payloads and the S&F principle, is being
developed and standardized in 3GPP Release 19 to address these challenges.

The main challenge of low-density LEO constellations lies in their discontinuous service and
feeder link, which disrupts the assumption of constant connectivity in mobile networks. Key
procedures like Attach/Detach, Tracking Area Update, data transmission, and Paging need
modifications due to signalling timers that control mobility and sessions, particularly for NAS
procedures. These adjustments must be completed within the limited visibility period of
satellites. To overcome these issues, a distributed 5G core network architecture compatible with
3GPP standards is introduced. It organizes core components into layers, integrating NTNs as a
key element. This architecture involves distributing core network functions as AMF, UPF, and
SMF between satellites (5G CN-SAT) and the ground (5G CN-GND), ensuring seamless
connectivity and efficient operations despite service interruptions as shown in Figure 4.2.

Dissemination level: Public | 43

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 4.2: Distributed 5G CN architecture between the satellite and the ground satellite
operation

4.1.2 UNIFIED MANO

The proposed MANO framework employs a hierarchical model for Management and
Orchestration (M&O) of services across ground, aerial, and space domains to tackle the scalability
issues. As shown in Figure 4.3, the M&O in the proposed MANO framework is structured into
three hierarchical levels. The E2E level, managed by the E2E Management & Orchestration
Component (EMOC), ensures global visibility and control over all domains and their
interconnections. At the domain level, each of the self-contained domains (including cloud, MEC,
RAN, and transport network) is managed by its own Domain Management & Orchestration
Component (DMOC) with other domain-specific functions. The infrastructure level comprises
Domain Infrastructure (DI) entities that provide an abstracted view of the underlying physical
and virtual resources.

Figure 4.3: High-level view of the proposed MANO with cloud, MEC, RAN and transport domain
separation

Dissemination level: Public | 44

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Furthermore, the Infrastructure Mobility Management (IMM) framework, a part of the

Supplementary Functions area, addresses the dynamic management needs of both static and non-
static DIs resources within 3D networks through a hierarchical structure.

Figure 4.4: High-level view of the proposed MANO with cloud, MEC, RAN and transport domain
separation

As illustrated in Figure 4.4, at the top, the Global Mobility Management Function (GMMF)
ensures efficient resource allocation and service continuity by registering and discovering
available domains as physical infrastructures move between them. The Domain Mobility
Management Function (DMMF) then takes over at a more localized level, managing domain-
specific mobility and facilitating seamless communication between the local and global mobility
management functions. At the bottom, the Local Mobility Management Function (LMMF) tracks
the locations and movement patterns of infrastructure components, providing real-time and
predictive updates for optimal resource management. This model allows the network to
dynamically adapt to changing conditions, ensuring robust service availability and performance
across various domains.

Moreover, to tackle the challenges of connectivity intermittence and consequently the need for
dynamic resource allocation in non-static network infrastructures such as satellites and high-
altitude platforms (HAPS), the concept of a Flexible Payload is introduced. This solution is
designed to allow the reconfiguration of onboard hardware, utilizing the Field-Programmable
Gate Array (FPGA) technology to adapt to changing service requirements and network conditions
dynamically. The Flexible Payload enables non-terrestrial systems (e.g., satellites) to function as
NFV Infrastructure (NFVI) nodes, managed by a Virtualized Infrastructure Manager (VIM),
enhancing the deployment of virtualized services. This framework supports the virtualization of
hardware boards' logical resources and software virtualization on the base operating system,
thereby transforming a satellite into a versatile NFVI that can dynamically host and manage
various services.

Dissemination level: Public | 45

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 4.5: Representative case of the proposed MANO architecture that includes mobility and
geo-localization management functions

The proposed MANO also addresses the complex challenges of managing and optimizing the
trajectories and geographical distribution of non-static network elements, such as satellites,
through the integration of a Geographic Information System (GIS). This GIS-based mobility plays
a critical role in planning and optimizing the movement paths and spatial distribution of these
elements. It also enables NTN operators to effectively simulate communication scenarios and plan
orbits, which are integral to managing the mobility of these dynamic infrastructures. The
integration of GIS capabilities into the unified MANO framework (Figure 4.5) significantly
advances its resource allocation, enhances coverage pattern accuracy, and optimizes network
performance. This development effectively streamlines operations and ensures consistent
service continuity amid dynamic geographical and environmental changes.

Additionally, the infrastructure management framework of the proposed MANO follows a

similar approach than the 3GPP Management Plane stack. Specifically, thanks to its hierarchical
structure, it can directly and natively be interfaced with the stack as depicted in Figure 4.6. In
particular, the 3GPP Management Plane functions are interfaced with corresponding Global,
Domain and Local Mobility Management Functions instances to receive the supplementary
information used to associate functional instances of the 3GPP framework with their location in
the 3D space. Consequently, it is possible to feed the 5G network control and management
algorithms with data potentially important for, e.g., mechanisms for handling dynamic network
topology, UE handovers, etc.

Dissemination level: Public | 46

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

3GPP
Management
Plane
CSMF
GMMF
NSMF

NSSMF DMMF

NFMF LMMF

Figure 4.6: 3GPP 5G System Management Plane in the proposed framework and its
interconnection with the proposed Infrastructure Mobility Management stack

4.1.3 E2E CONTROL OF UPF BASED ON MULTI -DOMAIN SDN

In the SDN-based UPF, the Control Plane SMF partner behind the N4 interface is the SDN
Control Plane in which the Packet Forwarding Control Protocol (PFCP) will be translated to the
SDN control protocols, e.g., Open Flow. The SDN Control Plane will be seen by SMF as the control
execution enabler of the fundamental packet routing and forwarding functionality of UPF
[3GPP24-23501], logically embedded within UPF. In the inherently multi-domain proposed
System infrastructure, the SDN-based UPF can be stretched over multiple SDN domains, thus,
needing the E2E inter-domain coordination at the Global Level (E2E Network Orchestrator with
Master SDN Orchestrator). Then, at the Administrative Domain Level, where the SDN
Orchestrator is located, providing the intra-domain focus and managing the SDN
Controller/WIM). It is assumed that UPF can be implemented as a set of dedicated
virtual/containerised/physical Network Functions, i.e., virtualised or physical SDN-enabled
Network Elements, SDN Controllers and necessary support functions (in the form of SDN
applications) needed to terminate N4 interface and support 3GPP functionalities. It is assumed
that the proposed MANO can deploy the UPFs via dynamic orchestration of virtualised
components (SDN Controllers, switches) or reuse the already existing ones (e.g., SDN Controllers
belonging to the proposed MANO Transport domains and physical SDN switches), ensuring
appropriate resource allocation and control privileges over the Control Plane/Data Plane devices.
In both cases, the E2E path setup process and session configuration are conducted via the N4
interface following standard 3GPP procedures.

Dissemination level: Public | 47

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

4.1.4 DIRECT HANDHELD ACCESS FOR UE

The 5G NR supports two waveforms, namely Cyclic Prefix OFDM (CP-OFDM) and Discrete
Fourier Transform-spread OFDM (DFT-s-OFDM). CP-OFDM is only for the downlink, whereas
both can be selected for the uplink. Regarding the DFT-S-OFDM, it may be selected based on the
channel conditions in the serving cell. Then, the gNB selects the waveform to be used for the
physical uplink shared channel and instructs the UE accordingly. In the cases of NTNs, both
waveforms have been adopted. However, the performance might be highly deteriorated in
scenarios with high Doppler shifts that may arise as the access point of a UE switches from a
terrestrial gNB to an NTN one. That is why OTFS is proposed as an alternative to OFDM-based
waveforms in NTN-related scenarios of high mobility of the NTN platforms, such as LEO satellites.
In such a case, through measurements of the Doppler shift by the UE and reporting to the gNB, a
threshold-based decision can be taken for switching to OTFS or not (based on the Doppler shift).
Such measurements can be performed under the assumption that the UEs are equipped with a
GNSS receiver, so that they know their position, and the satellites transmit their ephemeris data,
which contain their position and velocity. In cases, though, where the GNSS signals are weak and
the position of the UE cannot be estimated with accuracy, there might be significant residual
Doppler effects in the compensation process for OFDM-based waveforms. That is why operating
in the delay-doppler domain, used in OTFS, is advantageous because the channel becomes sparser
and varies on a much larger time fashion than in the time-frequency domain. There are several
open areas for investigation regarding the introduction of a new waveform, such as OTFS. These
include the design of synchronisation algorithms, random access protocols, and reference
symbols for the Doppler shift estimation.

Finally, regarding distributed simultaneous transmission from multiple satellites, which is

another important feature studied for achieving direct handheld access, the concept is like the
standardised coordinated multipoint (CoMP) aspect of LTE-Advanced, which allows joint
transmission to a UE from several distributed antennas. Here, the different satellites take the role
of the distributed flying antenna array for which synchronisation in time, frequency, and phase
needs to be done.

Dissemination level: Public | 48

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 4.7: Store and Forward mechanism in the distributed 5G CN architecture

The proposed solution for Store and Forward Satellite operations divides the 5G CN functions
between satellites (5G CN-SAT) and the ground (5G CN-GND). The satellite (5G CN-SAT) hosts
termination endpoints for N2/N3 and NAS protocols, while N8, N4, N6, and Nnef endpoints
remain on the ground (5G CN-GND). This architecture also supports multi-satellite scenarios,
allowing a single 5G CN-GND instance to interact with multiple satellites. The interface between
5G CN-SAT and 5G CN-GND is flexible, adapting to different constellations and network setups. In
one implementation, the core functionalities split equips satellites with gNB radio access
capabilities and necessary components to complete data and signalling exchanges within the
satellite's visibility period. The AMF positioned on the satellites enables NAS procedures within
the limited timeframe when the satellite is in contact with the user equipment (UE) as shown in
Figure 4.7.

In Store and Forward operation, the core network needs to inform UEs when the satellite
operates in the Store and Forward Satellite mode to avoid requests for unsupported services. A
satellite cell may switch between Store and Forward and normal modes based on network
policies. For example, if the satellite has simultaneous ground network connectivity, it may
choose to operate in either mode depending on the situation.

4.1.5 DISTRIBUTED NG-RAN FOR 5G/6G UNIFIED NTN NETWORKS

Network architectures exploiting functional split allow to implement flexible and scalable
solutions based on the principles of Network Function Virtualisation (NFV) and Software Defined

Dissemination level: Public | 49

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Networks (SDN) [ITUR23-M2160]. These allow to tailor the system to the requested use cases
and vertical services, and the corresponding Quality of Service (QoS), in addition to an improved
end-to-end network management and orchestration. In the context of Non-Terrestrial Networks
(NTN), such solutions are particularly promising also to enable the deployment of less complex
satellites that carry only the lower layers of the Control Plane (CP) and User Plane (UP) protocol
stacks, while leaving the higher layers either on-ground at the gateway side or on-board more
complex NTN platforms on the same orbit or higher. The former solution is more oriented to 5G-
Advanced NTN, while the latter is a solution that might be feasible for 6G NTN systems. In the
framework of 3GPP New Radio (NR) specifications, as discussed below, only option 2 is fully
enabled, i.e., a split in which the gNB Distributed Unit (gNB-DU) implements up to the Radio Link
Control (RLC)/IP layers. This solution might be particularly challenging in Non-Geosynchronous
Orbit (NGSO) NTN implementations due to the payload movement; in fact, the gNB-DU and the
Centralised Unit (CU) are connected via the F1 Air Interface that is persistent, i.e., it cannot be
closed and re-established without dropping all of the currently active Packet Data Unit (PDU)
sessions serving the User Equipment (UEs), [Nt+24].

Figure 4.8: F1 persistency in distributed NG-RAN for 5G/6G unified NTN networks

With functional split, the gNB-CU is in charge of managing the UE context and requests the
gNB-DU to allocate/modify the radio resources for that user. The radio resources are then
managed by the gNB-DU based on their availability. As such, the CU and DU are always belonging
to the same gNB when the F1 interface connecting them is established. In an NTN scenario, as
soon as the NTN node goes beyond the visibility of the serving gateway, the gNB-DU would
disconnect from its gNB-CU and, thus, beak the gNB requiring the creation of a new one resulting

Dissemination level: Public | 50

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

in the interruption of all its connections, as shown in Figure 4.8. It is worthwhile highlighting that
this is an issue specific to the F1 interface; in fact, handovers at NG interface level (i.e., between
the gNB-CU and the 5G Core) are allowed thanks to NG-flex configurations, i.e., each RAN node is
connected to all the Access and Mobility management Functions (AMFs) within an AMF region. In
this context, modifications to the F1 interface procedures or to baseline architectures with one
gNB-DU on-board shall be adapted to NGSO scenarios.

Potential solutions to the F1 persistency problem are based on multiple gNB-CUs or higher
layer handover procedures:

• Solution 1a: the single gNB-DU on-board could be connected to multiple on-ground gNB-
CUs for resiliency purposes; as soon as one becomes unavailable due to the feeder link
interruption, the second one can carry the F1 interface and its related traffic. However,
for this solution to be effective, it is required that the on-board gNB-DU is always in the
visibility of at least two gNB-CUs, i.e., two feeder links. This might be challenging due to
the very large number of required gateways and traffic overhead on the feeder links.
Moreover, the on-ground gNB-CUs shall be connected via Xn. A possible approach to relax
the first shortcoming is to also rely on Inter-Satellite Links (ISLs).

• Solution 1b: again, assuming the availability of multiple feeder links for the NTN node, it
might also be possible to implement intra-gNB-CU handover procedures as described in
[STK+24]. In this case, a single gNB-CU on-ground shall be connected to multiple gateways
and their corresponding feeder links.

• Solution 2: assuming two separate gNB-CUs on-ground, connected to two different on-
board gNB-DUs, it might be possible to implement the inter-gNB handover procedure
involving gNB-CU-UP change, described in [3]. In this procedure, it is clearly stated that
the F1 UE context is modified when sending the handover command to the UE, which also
indicates to stop the data transmission for the UE. Once the new gNB-CU-CP is initiated,
the radio bearer context modification allows it to retrieve the PDCP status and exchange
data forwarding information allowing the completion of the procedure (i.e., to connect the
UE to the new DU). Since this procedure involves the interruption of the PDU session and
its retrieval, it might be suited for application that support discontinuous data
transmission/reception.

All of the above potential solutions require detailed analyses in terms of architecture
feasibility, and 3GPP functions and procedures to be potentially modified.

Dissemination level: Public | 51

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

4.2 MULTI-CONNECTIVITY FOR 6G UBIQUITOUS

COVERAGE

4.2.1 6G MULTICONNECTIVITY

In 4G and 5G both Dual Connectivity (DC) and Carrier Aggregation (CA) were used as methods
to improve performance of the spectrum resources in the network. The master node could be on
a low frequency band while the secondary node may be a high frequency band cell, and hence
with worse coverage. The aim was to improve the usage of the network resources. However, in
practice this turned out to be more difficult, for the DC solution. One drawback with DC (and with
EN-DC) was that the master node (i.e., where the connection is terminated) will not have the most
recent information about the secondary node performance since the backhaul (Xn) connection
between the nodes may be too slow compared to the time scales of the variations in the radio
channel. The communication protocol (i.e., the flow control [3GPP22-38.420], [3GPP24-38.300])
between the master and secondary node estimates the throughput based on the
acknowledgements it receives from the secondary node. In some cases, if for example the
secondary node is a cell with high frequency, the coverage may drop quickly and cause long
packet delays for the connection over the secondary node. The master node may be unaware of
the drastically decreased performance and still send data to the secondary node over the Xn.
Another feature of DC is that DL and UL are always coupled and since the secondary connection
almost always have worse UL coverage than the master (i.e., the difference may very well be of
several dBs, depending on the frequency range), the secondary node feedback may become so
bad that this may cause a sharp increase in the round-trip times (or even a timeout), that in turn
would result in a decrease in the TCP/IP connection throughput.

Subnetworks

Traditional networks may not be able to efficiently handle the increasing number as well as
diversity of devices and applications. Additionally, 6G will introduce requirements for increased
coverage, lower power consumption, higher data rates, increased resilience, and increased
trustworthiness / user privacy, compared to 5G. Subnetworks are formed voluntarily by UEs
based on mutual trust and they may aid in achieving these KPIs/KVIs by offloading functionalities
from one node to another, by managing the radio resources more efficiently based on information
shared by the nodes and/or by providing connectivity to devices that are not in network coverage.

6G multi-connectivity

Therefore, to improve the MC solution for 6G, as well as simplify the solution by reducing the
number of architecture options, one option is to only allow MC between 6G enabled base stations

Dissemination level: Public | 52

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

and using one type of solution only. This solution should bring the best features from both DC and
CA, i.e. a CA/DC evolution. However, it seems natural to base the new 6G MC on the current CA
solution in 5G and improve it for 6G. The main reason for using CA as a base is the better UL
coverage of CA since the DL and UL connections are not coupled in CA. I.e., the best UL (i.e., the
PCell) can be used for UL response, which means that the UL coverage is often better for CA
compared to DC, as the UE does not have to split its limited uplink transmit power between two
concurrent UL connections. The new CA/DC evolution aims to decouple Downlink (DL) and
Uplink (UL) (e.g., two DL connections and one UL connection, see Figure 4.9) and inherent use of
in-active connections. For the in-active connections, the UE only need to sparsely monitor the
control signalling from the network. In addition, the in-active connections should be able to be
activated on a short notice. To increase the robustness of the system, there is a need for a more
flexible use of the UL so that the SCell may take over the role of control signalling in the UL.

Figure 4.9: Proposed 6G multi-connectivity solutions overview [HEX223-D33]

As part of the CA/DC evolution, faster addition of cells compared to 5G would be beneficial.
Further on, an enhanced mechanism for PSCell/SCell addition when transitioning from Idle mode
to Connected mode could be introduced. With this mechanism, the UE may perform
measurements during Idle mode of specific, pre-configured PSCells and not on all frequency
layers to avoid higher battery consumption.

Subnetworks architecture and solution

Forming inherently trustworthy subnetworks will also both extend coverage and create a
seamless communication system. To achieve the latter, a device may smoothly transition from
being served directly by a Base Station (BS) to being served by a Management Node (MgtN) and
vice versa. The MgtN is a UE which acts as the subnetwork’s primary node, being able to
communicate with the BS and other UEs. As an architectural option, the Control Plane (CP)

Dissemination level: Public | 53

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

entities of local devices can be flexibly deployed on the MgtN, allowing the subnetwork to use a
new lightweight SubNW CP (snCP) between the MgtN and the UE. In addition to relaying the UEs’
UP data to and from the overlay network, the subnetwork and especially the MgtN may assist a
UE with multiple CP procedures, such as RRC configuration, mobility, and Idle mode procedures.
Such an architecture is illustrated in Figure 4.10, where the snCP is used between the MgtN and
the UE. Note that the snCP is transparent to the NW, since it includes configuration and
procedures that take place within the SubNW. The content of the configuration, however, is still
managed by the BS.

Legend:
Virtual Connection
Physical Connection
Control Plane
UE3 Subnetwork Control Plane
CP
Control Plane Offloading
RAN-BS

MgtN

UE3
snCP
CP

snCP snCP CP CP

snUP L2 L2

UE2 UE3
UE1 PHY PHY PHY PHY
UE3
snCP
CP
UE3 Local Link, MgtN Access Network RAN-BS
e.g. Wifi

subnetwork

Figure 4.10: 1UE CP deployed at the MgtN and use of snCP within the subnetwork

4.2.2 NTN INTEGRATION FOR 6G MULTICONNECTI VITY

The unification of mobile network and NTN segments in 6G opens the door to a more effective
exploitation of the network resources available from terrestrial and non-terrestrial
infrastructures, but it also poses formidable challenges in the optimal exploitation of multiple
links because of the volatile nature of satellite links and the time-varying network topology,
whereby important traffic fluctuations, network handovers, etc., are dominant effects. These
important network oscillations may severely affect the overall system performance in terms of
overall quality of service and experience standpoint, whereby a well-defined network
architecture able to dynamically react and adapt to such changes is necessary for achieving an
effective network unification concept. In this respect, on the one hand the consolidated
framework of ATSSS represents a viable starting point, but its actual exploitation for scenarios
also including 3GPP-native NTN segments is not completely understood nor developed. Further

Dissemination level: Public | 54

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

to this, the optimization of link splitting, steering, and switching is a complex problem, so that
availability of AI-based solutions is the key to ensure an effective usage of the available network
resources. However, the placement of such AI functions in the considered unified NTN-6G
network architecture is still an open problem and their interactions with the overall ecosystem
needs additional studies to come up with a well-defined and solid architecture.

It is understood that ATSSS was initially conceived to allow native and non-native 3GPP
network segments to serve as access networks towards the 5G core network. In this respect, the
exploitation of multi-path protocol solutions is considered as particularly appealing because of
the intrinsic functionalities of traffic fetching and dispatching to different links, by even
simultaneously exploiting the resources available from more than one link and hence naturally
achieve load-balancing. In this framework, natural solutions are then represented by MPTCP
[RFC8684] and MPQUIC [IETF25-QUIC], the former being long standardized in IETF whereas the
latter is still in the approval process that may complete not earlier than late 2024. Despite this
late standardization planned achievement, MPQUIC offers interesting capabilities mostly
inherited from the characteristics of QUIC, especially for what concerns the establishment of
secure end-to-end data transactions, whereas MPTPC (similarly to TCP alone) must rely on
additional protocols or related extensions to provide security features.

On the other hand, the adaptive allocation to traffic flows to the different available network
segments necessitate advanced scheduling and fetching mechanisms, suitably coordinate by an
end-to-end controller. This opens the door to exploiting AI/ML-based algorithms for optimally
selecting the best paths through a dedicated path manager and make use of the recorded network
status (i.e. registered KPI, traffic fluctuations, etc.) through the NWDAF module as input to the
actual optimization modules. The general architecture going to be implemented in a dedicated
testbed in then sketched hereafter, where all the nodes and related protocol functionalities are
clearly identified. It is also worth noticing that AI functionalities are also attractive for
implementing effective traffic prediction algorithms, necessary to forecast possible variations in
demands of network resources and hence accordingly trigger the decision-making process at the
path manager for what concerns the adaptation or reinforcement of given traffic splitting and
switching policies

4.3 CONFLUENT TRANSPORT NETWORK

The current vision for 6G includes leveraging new spectral bands, such as high-frequency
millimetre-wave and terahertz ranges, to achieve higher peak data rates and utilizing advanced
technologies like ultra-large antenna arrays and cell-free (CF) architectures—enabled by CF-
mMIMO—to enhance spectral and energy efficiency, while supporting a higher number of

Dissemination level: Public | 55

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

simultaneous connections. Sensing at various network segments is also considered an inherent

capability to be developed in future 6G networks. At the same time, focus is put on enhancing
network orchestration capabilities by adding intelligence at network control and management
layers, aiming at optimizing performance and sustainability.

A proposed approach [ECO-eNET] is to integrate advanced technologies at the transport

network layer with high-capacity CF-mMIMO structures, advanced Open-Radio Access Network
(O-RAN) access nodes, and an intelligent control and network management plane. Sensing data
across the network infrastructure will be fused to the management plane to enable confluent-
mesh network structures delivering high availability and performance 6G network deployments.
These solutions can pave the way towards new service opportunities and features, transforming
the 6G ecosystem and advancing its alignment with the IMT-2030 goals and future user demands
[ITUR23-M2160].

Developing high-capacity wireless optical and radio fixed wireless (FWA) technologies and
their seamless integration referred to as ‘confluence’ with wired packet optical networks using
optical spectrum as a service (OSaaS), is the approach to introduce confluent fixed and wireless
optical and radio fixed wireless technologies for 6G front-/mid-/back-haul (xhaul) networks to
form cell-free mesh physical layer edge networks as illustrated in Figure 4.11 [Raj+24].

Confluent transmission makes use of a combination of RFW (at THz and sub-THz frequencies),
free space optical (FSO), switched flex grid wavelength division multiplexed (Flex-WDM) and
OSaaS fibre transmission capabilities to form mesh networks offering a flexible management of
high-capacity traffic with low latency and high energy-efficiency. In particular, RFW and FSO links
enable the formation of mesh networks at the edge, where deployment costs and complexity
prohibit wired mesh networks. These wireless links offer the efficiency and latency benefits of
mesh data transport, while they can also be used to transmit control plane signals to manage both
the wireless and wireline networks. This provides a new degree of freedom in the network
control that is exploited to facilitate wireline switching and low latency. Analog radio-over-fibre
(aRoF) transmitted over the wired network will efficiently be converted to RFW signals using
novel plasmonic devices. Analog signals will be multiplexed with digital signals throughout the
confluent xhaul network using a combination of optical and electronic switching to enable highly
dynamic power and spectrum management. The efficiency of using such confluent xhaul network
with high-density cell-free radio access networks using coordinated multipoint and distributed
multiple input multiple output (d-MIMO) techniques needs to be investigated for delivering high
data capacity over a wide range of spectral bands, including line-of-sight (LoS) communications
subject to severe blocking and fading challenges.

Dissemination level: Public | 56

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

The potential of technologies such as RFW (at THz and sub-THz frequencies), FSO, and Flex-
WDM fibre optics to deliver required performance by 6G need to be examined. Electronic RFW
transceivers will be replaced by plasmonic-based ones to extend the reach of the RFW link and
significantly reduce the energy consumption of the network. Moreover, creating a transparent
fibre-FSO interface supporting multiple modulation formats would be desirable. This could be
possible by novel, low-cost, adaptive photonic components and lantern technology. Optically
switched and modulation format adaptive optical spectrum as a service (OSaaS) can be used to
efficiently multiplex and transport the signals from these diverse transmission links in mesh
configurations and connect them with edge computing resources. OSaaS will also enable fibre and
radio sensing signals to be carried alongside the communication signals.

Figure 4.11: Confluent mesh networking

The performance of the physical layer technologies and their potential to deliver improved
capacity, energy efficiency, and latency depends strongly on their control and management. This
however imposes several challenges as radio and transport networks are currently separately
engineered and controlled introducing the need for 5G core, transport and radio access network
control to evolve towards increased coordination and compatibility. AI-assisted network control
can facilitate optimised control and management capabilities across wireless and wireline
systems. Therefore, building tools that can exploit the confluent networking capabilities within
this evolving and increasingly integrated control environment is crucial.

A flexible and scalable control framework with extendable Application Programming

Interfaces (API) on top of the B5G/6G RAN, the Core, and the transport network, can facilitate
monitoring and programmability of the underlying network infrastructure. This will enable an
end-to-end platform that can support customised service delivery in response to the service
requirements in the most resource and energy-efficient manner. The monitoring data collection
ensures appropriate system initialization and allows continuous optimisation of the entire

Dissemination level: Public | 57

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

system operation. Such a monitoring system can subscribe and collect both high-level (E2E
service related such as throughput, packet latency, jitter, etc.) and low-level statistics
(network/compute resource utilisation, bit error rate, packet error rate, power consumption,
physical layer characteristics of the RFW links and optical switching nodes). These statistics can
be exposed to other NFs, such as the Network Data Analytics Functions (NWDAF), to provide
recommendation services. Such system capabilities enable lower layer decisions such as optimal
mapping of the wireless domain service characteristics to the optical transport parameter
configuration. An integrated control plane development, supporting the seamless cross-domain
service delivery through the wireless and the transport network domains, will be beneficial.

4.4 REFERENCES
[3GPP22-38.420] 3GPP TS 38.420: "NG-RAN; Xn general aspects and principles", V17.2.0, September 2022
[3GPP24-23.501] 3GPP, “System architecture for the 5G System (5GS)”, 3rd Generation Partnership Project,
Technical Specification TS 23.501, ver. 19.1.0, September 2024. [Online].
https://portal.3gpp.org/desktopmodules/Specifications/SpecificationDetails.aspx?specificationId=3144.
[3GPP24-38.300] 3GPP TS 38.300 V18.4.0: “NR and NG-RAN Overall Description”, V18.4.0, December 2024
[5GSTAR24-D52] 5G-STARDUST Deliverable D5.2, “Preliminary Report on Multi-Connectivity and Software
Defined Network Control”. Online. https://www.5g-stardust.eu/wp-
content/uploads/sites/97/2024/08/5G-STARDUST_D5.2_1.0.F.pdf June 2024.
[6GNTN24-D35] 6G-NTN Deliverable D3.5, “Report on 3D multi layered NTN architecture (2nd version),”
March 2024.
[BGG+23] Bahare, M. K., Gavras, A., Gramaglia, M., Cosmas, J., Li, X., Bulakci, Ö., Rahman, A., Kostopoulos, A.,
Mesodiakaki, A., Tsolkas, D., Ericson, M., Boldi, M., Uusitalo, M., Ghoraishi, M., & Rugeland, P. (2023). The 6G
Architecture Landscape - European perspective. Zenodo. https://doi.org/10.5281/zenodo.7313232
[BLG+23] Ömer Bulakçı (ed.), Xi Li (ed.), Marco Gramaglia (ed.), Anastasius Gavras (ed.), Mikko Uusitalo
(ed.), Patrik Rugeland (ed.), Mauro Boldi (ed.) (2023), "Towards Sustainable and Trustworthy 6G:
Challenges, Enablers, and Architectural Design", Boston-Delft: now publishers,
http://dx.doi.org/10.1561/9781638282396
[ECO-eNET] Project ECO-eNET: Developing an intelligent network of technology enablers to seamlessly
connect the human, physical, and digital worlds. Online. https://www.eco-enet.eu/
[HEX223-D33] HEXA-X-II Deliverable D3.3, “Initial analysis of architectural enablers and framework”
https://hexa-x-ii.eu/wp-content/uploads/2024/04/Hexa-X-II_D3.3_v1.0.pdf, April 30, 2024
[IETF25-QUIC] “Multipath Extension for QUIC", draft-ietf-quic-multipath-10, IETF, Expiration date: 9
January 2025
[ITUR23-M2160] ITU-R, “Framework and overall objectives of the future development of IMT for 2030 and
beyond”, M.2160, November 2023. [Online]. Available: https://www.itu.int/rec/R-REC-M.2160/en.
[KCC+22] T. Kellermann, R. P. Centelles, D. Camps-Mur, R. Ferrús, M. Guadalupi and A. C. Augé, "Novel
Architecture for Cellular IoT in Future Non-Terrestrial Networks: Store and Forward Adaptations for
Enabling Discontinuous Feeder Link Operation," in IEEE Access, vol. 10, pp. 68922-68936, 2022, doi:
10.1109/ACCESS.2022.3184720.
[Nto+24] K. Ntontin et al., "ETHER: A 6G Architectural Framework for 3D Multi-Layered Networks," 2024
IEEE Wireless Communications and Networking Conference (WCNC), Dubai, United Arab Emirates, 2024,
pp. 1-6, doi: 10.1109/WCNC57260.2024.10570731
[Raj+24] R. Raj et al., "Towards Efficient Confluent Edge Networks," 2024 Joint European Conference on
Networks and Communications & 6G Summit (EuCNC/6G Summit), Antwerp, Belgium, 2024, pp. 1163-

Dissemination level: Public | 58

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

1168, doi: 10.1109/EuCNC/6GSummit60053.2024.10597093.

[RFC8684] RFC 8684, "TCP Extensions for Multipath Operation with Multiple Addresses", IETF, March 2020
[STK+24] M. M. Saad, M. A. Tariq, M. T. R. Khan and D. Kim, "Non-Terrestrial Networks: An Overview of
3GPP Release 17 & 18," in IEEE Internet of Things Magazine, vol. 7, no. 1, pp. 20-26, January 2024, doi:
10.1109/IOTM.001.2300154.

Dissemination level: Public | 59

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

5 ARTIFICIAL INTELLIGENCE AND

COMMUNICATIONS
The 6G architecture must inherently support AI as a Service (AIaaS) for both internal network
operations and external applications, thereby enhancing automation and distributed intelligence.
A critical application of AI in this context is network automation. The primary challenges to AI
adoption for network automation involve the sustainability of the machine learning (ML) training
process, particularly in terms of energy consumption and environmental impact, as well as the
perceived trustworthiness of AI decisions. To address these challenges, ML training and inference
should employ techniques and architectures that optimize computing resource utilization and
power consumption. This includes selecting the optimal placement of AI/ML functions based on
data location, utilizing computing nodes powered by renewable energy, adopting collaborative
learning techniques to distribute the training process, and properly scheduling training and re-
training activities. Regarding trustworthiness, explainable AI techniques should be employed to
elucidate the rationale behind predictions and automation actions, mitigating potential biases
and possibly retaining human oversight. Furthermore, ensuring robustness against security
attacks and maintaining the privacy and confidentiality of training data in distributed AI systems
are imperative. Federated learning could address these privacy concerns by sharing only the
trained models rather than the raw data. Consequently, 6G architecture should emphasize
ubiquitous intelligence, sustainability, and security/resilience.

This chapter is structured as follows. Section 5.1 presents several options on how to introduce
the AI/ML framework to the 6G architecture, including RAN, CN and management. Section 5.2 is
then discussing how to use intents. Section 5.2 discusses means to introduce intent-handling
function that comprehends sophisticated and abstract semantics. Section 5.3 then addresses the
management and orchestration using AI. Such as programmability, pervasive monitoring, and
distributed orchestration etc. Section 5.4 then describes AI solutions for the digital twin use case
for developing a virtual environment where e.g., AI-driven network functions can be developed
and tested.

5.1 AI/ML FRAMEWORK / INTELLIGENT PLANE

This section discusses several options on how to introduce the AI/ML framework or the
intelligent/data plane as it is also called.

Section 5.1.1 presents several options on how to introduce the AI/ML framework to the 6G
architecture. It also outlines the necessary functions or enablers for this and outlines a possible

Dissemination level: Public | 60

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

framework. In section 5.1.2 the intelligent plane is introduced, using O-RAN architecture as a
basis. Cloud nativeness is an important aspect of 6G, and how to improve the data plane (user
plane) to be more cloud friendly is handled in section 5.1.3. Finally, section 5.1.4 deals with the
RAN aspects i.e. using AI to improve waveforms, transceivers and protocols.

5.1.1 AI/ML ENABLERS AND FRAMEWORK

The AI enablers within the 6G data-driven architecture encompass architectural means and
protocols, Machine Learning Operations (MLOps), Data Operations (DataOps), AI as a Service
(AIaaS), and intent-based management [HEX223-D33]. These AI enablers constitute a robust
framework that seamlessly integrates AI into the compute continuum of 6G networks, facilitating
advanced automation and distributed intelligence.

MLOps is focused on operationalizing machine learning models by ensuring their smooth

deployment, version control, and continuous monitoring within the overarching architecture.
The architecture requirements for MLOps include access to high-quality data, scalable data
storage solutions, computational resources for data processing and model training, and stringent
security and trust measures.

DataOps enables efficient data collection, integration, and management, providing MLOps with
timely and high-quality data. Essential architectural requirements for DataOps include robust
data quality management functionalities, end-to-end data pipelines that effectively serve MLOps,
and version control mechanisms for the collected data (refer to Figure 5.1).

Building upon MLOps, DataOps, and intent-based management, the AIaaS framework delivers
AI services across various network segments and to end-users. This framework necessitates the
development of new APIs for both internal network exposure and external end-user access.
Furthermore, AIaaS demands rigorous security measures and regulatory compliance, along with
feedback loops for continuous improvement and resource optimization.

In the context of 6G networks, the integration of AIaaS is pivotal for enhancing network
performance, reliability, and intelligence. It supports a wide array of applications, from real-time
analytics to predictive maintenance, thereby driving the evolution of next-generation network
services. By leveraging advanced AI/ML technologies and methodologies, the 6G architecture
aims to achieve unprecedented levels of automation, efficiency, and user experience, establishing
a foundation for future innovations in telecommunications.

Dissemination level: Public | 61

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 5.1: AI/ML framework [HEX224-D33]

5.1.2 INTELLIGENCE PLANE IN O-RAN

Intelligence Plane, depicted in Figure 5.2, which works as a cross-domain management entity,
integrating control and monitoring functions across RAN, Core and Edge domains and fostering
the creation of advanced ML models [BEG23-D41]. The proposed Intelligent Plane incorporates
an AI Engine, which provides a serverless execution environment hosting the AI/ML models,
offering inference and training services to the rApps/xApps implementing the control loops by
following a loosely coupled approach.

The AI Engine hosts the ML models to offload inference tasks from the RICs and implement the
necessary AI/ML workflows and services. As shown in Figure 5.2, the AI Engine manages the
AI/ML pipelines, including model management, monitoring, training, serving, and a data lake with
prepared data. The models are served in a serverless way, which enables efficient scaling of
workloads in production. In the case of O-RAN, the inference of the models is exposed to the
control rApps/xApps though AIA1 and AIA2 interfaces plus associated AI Engine Assist
rApps/xApps, what allows to decouple the implementation of control-loops from the
management of ML models.

Dissemination level: Public | 62

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 5.2: Intelligence Plane

This facilitates model reusability by different control rApps/xApps and allows the integration
of AI/ML workflows through the AI Engine independent of the RICs implementation. A similar
approach could be adapted for Edge and Core domains, using AIA3 and AIA4 interfaces to expose
AI Engine AI/ML services to Edge applications or NWDAF analytics. In [BEG23-D41], this
approach is followed to develop energy efficiency optimizations, including the dynamic
management of Edge and O-Cloud compute resources, of relay nodes and of Reconfigurable
Intelligent Surfaces (RIS) via extended O-RAN interfaces.

5.1.3 CLOUD NATIVE DATA PLANE

The cloud native principle has widely been accepted by industry and applied to deploy
traditional computing applications, ensuring better management and utilization of computing
resources. Environments like a Kubernetes cluster uses proxy load balancers to ensure seamless
load distribution among instances and automatically handles on-demand up and down scaling.
This approach ensures high-level of flexibility in using computational resources and enables
application developers to focus on the business logic’s implementation while scalability,
performance and deployment questions are handled by the environment. Though cloud native

Dissemination level: Public | 63

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

approaches are applied in 5G systems to deploy control plane entities like AMF and SMF, it has
not been extended to the user plane yet where much stricter performance requirements of packet
processing logic need to be satisfied. Though the cloud native approach would also be beneficial
for user/data plane applications, additional challenges need to be solved to handle heterogeneous
programmable targets to be used for packet processing (e.g., programmable switches, CPUs,
DPUs, SmartNICs, IPUs, FPGA-based NICs). These challenges include: (i) hiding the
implementation and deployment details of the underlying network function data planes, (ii)
providing seamless and dynamic offloading and optimization in the data plane, i.e.,
hardware/software target selection, disaggregation, (iii) providing tools to better use the
resources of hardware data planes with network function isolation, e.g., by enabling the
deployment of multiple network function data planes on the same hardware, mimicking data
plane hardware virtualization. Current programmable data plane targets do not support multi-
tenant usage and virtualization.

Figure 5.3: High-Level Architecture and Innovations [DESIRE24-D22]

A new architectural component called infrastructure management layer (IML) was proposed
in [DESIRE24-D22] to separate concerns of the packet processing business logic and the
infrastructure layer (Figure 5.3). IML basically acts as a combination of a Virtualized
Infrastructure Manager (VIM) and a hardware abstraction (HAL) layer. IML is responsible for
managing a pool of resources (e.g., located on a 6G site).

In 6G network services, packet traffic needs to be forwarded through different network

functions (NFs). Each NF has a control plane and a data plane. IML focuses on the deployment and

Dissemination level: Public | 64

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

run-time management of data plane components. An NF data plane component implements the
packet processing logic and can be executed on various targets including smartNICs, ASICs,
FPGAs, IPUs and DPUs, in addition to traditional CPU resources. IML is responsible for selecting
the appropriate target(s) and number of instances to execute the NF data plane and configure the
virtual links between them at deployment time. Virtual links are created by infrastructure
network functions implementing traffic forwarding and routing between NFs. To enable run-time
optimization and hide the underlying optimization from the NF control plane, IML introduces a
control plane proxy using a common northbound API (e.g., P4Runtime [P4R24]) that provides a
single-instance view of the data plane component to the NF control plane. The proxy hides the
underlying data plane optimization like load balancing between multiple data plane instances of
the same NF data plane or offloading heavy hitter users to hardware data planes. To enable the
better utilization of data plane hardware resources, IML has a subcomponent called P4-MTAGG
[BKL+24a], [ BKL+24b] that is a compiler-based virtualization tool for P4 [BDG+14]
programmable hardware targets. It enables the deployment and execution of multiple P4
programs on the same P4 hardware in an isolated way. The control plane access to the different
data plane programs is also isolated by the IML’s control plane proxy component.

5.1.4 AI AIR INTERFACE

AI will be crucial in meeting the technical and societal needs of 6G communication systems,
enabling energy-efficient, user-centric communications. By leveraging advanced AI techniques—
such as reinforcement learning, transfer learning, and semantic communications— waveforms,
transceivers and protocols can be customized for diverse scenarios, devices, and users. In
contrast to 4G and 5G, 6G will benefit from vast datasets and over a decade of machine learning
progress [AHM24].

For example, future use cases like mission-critical video streaming in factories with 360° 4K
cameras demand extreme bandwidth and low latency, requiring customized physical layer (PHY)
and protocol designs. However, today's rigid architecture makes such customization too
expensive. AI can optimize waveforms, MIMO processing, and networking protocols for specific
devices and environments, offering performance and energy efficiency. This AI-driven approach
allows flexible, cost-effective communication systems, addressing challenges like rural coverage
and reducing reliance on rigid network architectures.

The AI-AI concept places the users’ communication needs and application-specific
requirements at the centre of the design as depicted in Figure 5.4. Then, tailor-made waveforms,
transceivers, signalling, protocols, and hardware implementations are optimized adaptively and
on-demand within a modular architecture to support these requirements.

Dissemination level: Public | 65

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 5.4: Artificial Intelligence native Air-Interface (AI-AI) architecture component

As shown in the AI-AI architecture in Figure 5.4, novel AI-solutions for the physical, MAC and
RRM layers must be developed to drive the concept. A few of such solutions are briefly described
below.

Neural network-based receiver with end-to-end learning [CENTR24-D21]: An AI-native

receiver will constitute a major L1 component of the AI-AI concept. The neural network-based
receiver depicted in Figure 5.5 replaces traditional receiver processing blocks including channel
estimation, equalization and demapping in 5G with a single NN which is trained in an end-to-end
version. The NRX is compatible with 5G NR PUSCH and can support pilotless communication and
custom constellation.

Figure 5.5: Neural network-based receiver

Dissemination level: Public | 66

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

5.2 INTENT-/GOALS-DRIVEN COMMUNICATIONS

This section deals with how higher-level languages (such as express intents for network
configuration or Semantic communication) can used to control the network. Section 5.2.1
describes how intents and the cognitive component can be used in the network. Section 5.2.2
describes how information exchange can be done by using semantic context of AI. Section 5.2.3
discusses an architecture to enable high security and privacy in future 6G networks using intent
based interfaces.

5.2.1 INTENT-DRIVEN COGNITIVE 6G NETWORK

The Cognitive Coordination [GKF+24], [ARF+24] component is an intent-handling function

that comprehends sophisticated and abstract trust intent semantics (divided into the five
trustworthiness taxonomies of Safety, Security, Privacy, Resilience, and Reliability), calculates the
ideal goal state, and organizes activities to transition the SAFE-6G system into this trustworthy
state. The function will be able to research possibilities about the applicability of the five functions
in providing the desired degree of trust, learn from precedents, and assess the feasibility of
actions based on their expected results.

Figure 5.6: SAFE-6G Functional architecture of the cognitive layer

Cognitive Coordination [SAFE6G] component is an autonomous service and network

operation by combining well-known AI technologies inside a flexible framework. The cognitive
layer, as shown in Figure 5.6, acts as an interface between tenants/users/3rd party apps and the

Dissemination level: Public | 67

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

network/environment via the 6G service exposure provider interface. The network architecture
in 6G in SAFE-6G will comprise a fully end-to-end Machine Learning (ML) and model accessibility
encompassing autonomic networking by taking advantage of AI/ML capabilities for instance
making use of supervised and unsupervised machine learning algorithms as well as
reinforcement learning.

AI/ML techniques will help to manage the edge-cloud continuum, dealing with the
heterogeneity of data source and the high number of tenants also, correlating data from far-edge,
edge, and Core Network towards providing the requested level of trust. Different ML algorithms
(e.g., supervised, unsupervised, federated or reinforcement learning) will be considered to
efficiently improve the coordination of resource and service orchestration, as well as for resource
monitoring by implementing orchestration actions (e.g., closed-loop automation, proactive
alerting, automated healing, and/or predictive NF scaling/placement, etc). AI will ensure a
proficient, optimal, and continuous end-to-end orchestration. Hence, the end-to-end
orchestration needs to span over different domains while having a coherent view of their own
scope and purpose. In such context, in SAFE-6G complex and distributed system will be
coordinated by designing cognitive and effective architectural interfaces to manage and
synchronize both, operations and communication.

The cognitive coordination component consists of three major components: a knowledge base,
a reasoning engine, and an agent architecture. The knowledge base includes an ontology of trust
intents as well as domain-specific knowledge, such as the current state of the system. The domain-
independent reasoning engine will use the knowledge graph as the primary coordinator function
for locating actions, assessing their impact, and ordering their execution in order to provide the
requested level of trust to the requested tenant/user and/or third-party application. Finally, the
agent design allows for the use of an unlimited number of models and services.

5.2.2 GOAL-ORIENTED AND SEMANTIC COMMUNICATION IN 6G AI -

NATIVE NETWORKS

AI's rapid rise impacts communication infrastructure by demanding vast data and
computational resources for training large models, primarily generated by edge wireless sensors.
This intensifies network pressure. Furthermore, advanced AI increases device intelligence,
shifting communication to information exchange within the semantic context of AI.

Currently, human-oriented traffic dominates networks. However, with the surge in connected
autonomous vehicles, smart wearables, robots, and AR/VR equipment, machine-type
communications will dominate, requiring different KPIs. For instance, video signals for machines
focus on task-relevant information, unlike those for human consumption, which must meet

Dissemination level: Public | 68

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

latency and quality requirements. This shift redefines communication networks from reliable bit
transmission infrastructures to networks of intelligence, blurring the lines between
communication, computation, and intelligence, following a goal-oriented design approach
[Str+24].

[Str+24] outlines contributions to the emergence of this new paradigm with a novel goal-
oriented communication architecture and enabling algorithms that make communication
converge with computation in a jointly data- and model-driven manner, such as:

Semantic Engine: will be positioned in the service management and orchestration unit of a
network, which is responsible for the efficient and effective delivery of semantic-oriented
services, through the orchestration of semantic information resources processing, semantic
model’s lifecycle and user experience management.

Semantic Radio Intelligent Controller (S-RIC): will be linked with the control and user
planes of O-CU for resource, connection and quantity of service management. S-RIC and O-DU
connectivity will also be supported for the disturbed semantic information processing ability. A
rich set of interfaces will be defined additionally within the O-RAN architecture for the above
connectivity. S-RIC will be designed as a programmable and extensible unit to facilitate the
deployment of diverse semantic applications. It will also support different service time
requirements, ranging from non-real-time to near-real-time and real-time, for these applications.

Application Plane: The application plane is orthogonal to the semantic plane in the diagram,
which provides the interfaces for the semantic applications across edge devices or user
equipment. Meanwhile, this plane will interact with the core network, near-real-time RIC, and
newly defined semantic interfaces in the O-RAN for holistic tasking scheduling and optimised
resource allocation in a security way.

Semantic-empowered UE and Edge: The UE and edge devices are required to be mounted
with computational and learning capabilities, for the initial semantic information extraction from
the raw image, video and sensor data. Meanwhile, the intelligent O-RU will also be explored for
better real-time semantic processing (including extraction and interpretation) capability.

Knowledge database: The functionality of the knowledge base is implicitly reflected in this
structure. It has not been explicitly depicted because it operates in almost every semantic
processing module. The training and validation of semantic models both require support from a
knowledge base, and a unified and consistent knowledge base is crucial for the successful
extraction and translation of semantic information by the model.

Integrate by design semantic and goal-oriented principles into the AI/ML architecture to
optimize control signalling necessary in the protocols and thus markedly reduce the

Dissemination level: Public | 69

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

communication overhead without sacrificing correct and timely operation of the protocols.
Examples include innovative L1/L2 protocols for massive access with targeted semantic content
selection. To this aim, the novel concept of a semantic RAN intelligent controller (S-RIC) is
introduced. The O-RAN architecture introduces the RAN Intelligence Controllers (RICs) as ML-
based functional blocks operating in a closed-loop scheme at non-real time (e.g., seconds) and
near-real time (e.g., tens of milliseconds) time granularity. The interaction among states and
messages at different time scales offer the opportunity for bringing in the semantic
representations and enable effective interaction among various control loops. This has a clear
potential to progress beyond the SoTA and introduce a novel real-time semantic control layer,
applicable to O-RAN, and design its functionalities and interfaces. Specifically, this building block
can directly interact with L1/L2 DU related tasks, leveraging the semantic communication
paradigm to boost the overall network efficiency and performance.

Figure 5.7: The conceptual architecture [Str+24]

5.2.3 6G ML TRAINING, INTENT BASED INTERFACE, NETWORK

DIGITAL TWIN

Security and privacy enhancement in future 6G networks can be a quite challenging and
demanding task, due to the vast number of potential threats and attacks and their diverse nature
compared to 5G networks (indeed, a larger attack surface is expected in 6G networks). In the
same context, the interconnection of a vast number of devices and the support of heterogeneous

Dissemination level: Public | 70

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

deployments (exploiting the cloud continuum paradigm), which are both key concepts of the 6G
era, might leverage security and privacy concerns, since not all devices will have the capability to
execute advanced security protocols due to their hardware constrained nature.

To enable high security and privacy in future 6G networks, artificial intelligence/machine

learning (AI/ML) approaches can be deployed. Compared to conventional non-ML detection
techniques, ML-based misbehaviour detection provides both a higher detection accuracy against
unknown zero-day attacks, as well as a reduced false detection rate.

To be efficient, the AI/ML needs to collect a vast amount of data from the network to train
models that can represent input/output pairs with minimum performance loss, can leverage
security and privacy mitigation via the extraction of abnormal data patterns and the enforcement
of the appropriate actions.

The deployment of ML approaches for threat detection and mitigation in the 6G landscape is
dictated by various key driving factors: i) computational efficiency of the deployed approaches,
ii) identification of multiple and even correlated threats and attacks, iii) continuous refinement
of the ML approaches and knowledge distillation, as well as iv) creation of multiple network
intents per case for network recovery [GNT+224].

Dissemination level: Public | 71

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 5.8: Proposed Architectural Approach for the DT integration.

In this context, artificial intelligence/machine learning (AI/ML) approaches that collect a vast
amount of data from the network to train models that can represent input/output pairs with
minimum performance loss, can leverage security and privacy mitigation via the extraction of
abnormal data patterns and the enforcement of the appropriate actions. Compared to
conventional non-ML detection techniques, ML-based misbehaviour detection provides both a
higher detection accuracy against unknown zero-day attacks, as well as a reduced false detection
rate.

The deployment of ML approaches for threat detection and mitigation in the 6G landscape is a
quite challenging procedure, dictated by various key driving factors: i) computational efficiency
of the deployed approaches, ii) identification of multiple and even correlated threats and attacks,
iii) continuous refinement of the ML approaches and knowledge distillation, as well as iv) creation
of multiple network intents per case for network recovery [GNT+224].

Dissemination level: Public | 72

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

The proposed reference architecture is illustrated in Figure 5.8. Key elements encompass the
ML/deep learning (DL) training components dedicated to threat detection, the intent-based
networking (IBN) components as well as the digital twin (DT) module.

ML/DL training (Distributed Threat Detection – DTD): This module is responsible for the
distributed ML training. To this end, privacy preserving solutions are leveraged, such as federated
learning [LST+20]. Before the actual training takes place, preprocessing and feature preparation
takes place. Trained ML models are stored in a local database, where they can be retrieved on
demand. All procedures are orchestrated by the machine learning function orchestrator (MLFO).

Intent-based Threat Mitigation (IBTM): Mitigation and preventive actions will be applied to
the network using an intent-based interface (IBI) to facilitate human awareness. An intent should
clearly define the desired state of the network while keeping its specification human-readable.

Sandbox – 6G Network DT: The network DT acts as a dynamic representation of the mobile
network, constantly learning and evolving alongside the real network environment. ML
algorithms, within this DT framework, can leverage historical data, network topologies, and user
behaviour patterns to model normal network behaviour and promptly identify deviations that
may indicate malicious activities. This integrated approach not only enhances the precision of
threat detection but also empowers security systems to both anticipate and proactively mitigate
potential risks as well as analyse the impact of any proactive action to be taken.

Within the DTD module, the data collected either directly from the network or from the
emulated context with the help of the digital twin representation, are used to train the
appropriate ML models for threat mitigation. In this context, the IBTM component interacts with
the DTD where the proposed intents, being outputs of the ML model training, are translated into
decisions to be applied into the network, as previously mentioned. Therefore, DTD provides a
high-level description of the mitigation or preventive actions to be enforced in the different 6G
components in the form of an intent. Upon receiving the intent, the intent-based threat mitigation
module (IBTM) module will match the received intent with existing information in the knowledge
base and will select the proper matching policies. After checking whether the results of policies
are as expected and if the new policies are aligned with existing policies and the decision to be
taken, the polices can be enforced in the 6G infrastructure. All actions and policies aim either at
strategically relocating virtual network functions (VNFs) to alternative cloud-native instances,
such as containers or other cloud hosts or completely isolating malicious nodes. Access
revocation to the 6G infrastructure is also supported to minimize potential risks and exposure to
the threat. All proactive measures are designed to effectively mitigate identified threats and
restore the optimal functionality of the 6G Core network.

Dissemination level: Public | 73

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

5.3 MANAGEMENT AND ORCHESTRATION

6G network management needs to integrate several technical enablers, addressing the
challenges of the multi-technology and distributed nature of future 6G infrastructures, the
diversity of services to be delivered and the variety of stakeholders contributing to the whole
ecosystem. Key features are the deep programmability and pervasive monitoring, which jointly
enable synergetic, distributed orchestration combined with higher levels of network automation
in scalable, multi-domain environments. Techniques like AI/ML algorithms, zero-touch closed
loops and network digital twins are applied to bring increasing intelligence in the network,
distributed through different layers and domains with functions deployed and configured on-
demand following cloud-native, Service Based Architecture (SBA) and as-a-Service patterns. The
usage of service intents mixed with controlled but powerful exposure of network capabilities
facilitates more effective interactions with verticals and digital service providers, a key aspect for
the monetization of added value network services, beyond mobile connectivity. Sustainability and
trustworthiness natively follow a pervasive and “by-design” integrated approach. This involves
the adoption of unified architectural principles and the embedding of algorithms, protocols, and
workflows for user-centric, energy-efficient, and security procedures. Notable, energy efficiency
and security are considered not only as primary objectives of provisioning and automation
decisions, but also as principles for the design and deployment of Management and Orchestration
(M&O) components, introducing e.g., elements for sustainable MLOps or Federated Learning for
privacy-preserving and explainable AI techniques.

5.3.1 EDGE CONTROL

Supporting edge computing applications (e.g., AI applications) is one of the most exciting
features of future mobile networks. These services involve collecting and processing voluminous
data streams right at the network edge to offer real-time services to users. However, their
widespread deployment is hampered by the energy cost they induce on the network.

Recently, O-RAN has studied collaboration and convergence across domains to enable cross-
domain AI optimization [ORAN23-nGRG]. As RAN virtualization enables the use of Commercial-
Off-The-Shelf (COTS) hardware for deploying RANs, it opens the door to the joint orchestration
and management of RAN, Core, and edge applications. The network's role in these services
extends beyond merely transmitting and processing data in transit. Instead, the network must
directly enhance service performance by optimizing for accuracy (reliable inferences), end-to-
end latency (swift inferences), and task throughput (inferences per second) in a resource-efficient
manner. This last requirement is critical because these services generate substantial data flows,
involve intensive computations, and consume significant energy.

Dissemination level: Public | 74

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

To this end, the Intelligence Plane [BEG24-D42] includes an Edge Control function in the SMO
conceived to control the resources of edge servers hosting edge applications dynamically. This
function is exposed to the non-RT RIC whose rApps aim at optimizing edge resources jointly with
radio resources to enhance energy efficiency. It also enables the creation of control loops focused
on joint optimizations. Since this approach has several similarities with the O-cloud management
performed by the SMO, an interface leveraging O2, denoted as O2+, has been specified between
the SMO and the edge server. Through this interface, the SMO shall be able to obtain the
configuration of the edge, monitor its resources, and apply configurations related to energy-
saving modes, CPU frequency control, CPU allocation, or GPU allocation.

Functions for optimising edge applications and RAN configuration policies are deployed as
rApps in the O-RAN's non-RT RIC to enforce radio control policies in O-RAN-compliant eNBs or
gNB. The edge control rApps interact with O-RAN's A1 interface (specifically, the A1's Policy
Management Service) to enforce the corresponding radio policies. An xApp handles the A1 service
from O-RAN's Near-RT RIC side and uses an E2 interface to forward radio policies to the Base
Station. The E2 interface is also used to gather BS KPIs, which are forwarded to the non-RT RIC
through the O1 interface. Then, a second xApp manages data KPIs received from the vBS and
sends them to the Data Lake. Figure 5.9 summarizes the architecture of this use case and the
involved interfaces.

Figure 5.9: Intelligence Plane architecture [BEG24-D42]

5.3.2 6G NETWORK MANAGEMENT AND AUTOMATION

The overall M&O framework proposes three main features, supported transversally by several
technical enablers: intent-based service management, synergetic orchestration in the computing
continuum and cognitive closed loops for network automation at runtime.

Dissemination level: Public | 75

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

The synergetic orchestration enables the management of network services over

programmable resources in the computing continuum, using techniques for distributed, de-
centralized and/or federated management of the available resources. The scalable management
of resources that span across the computing continuum (IoT devices, extreme edge devices,
edge/cloud infrastructure, Figure 5.10) is crucial, leading to solutions based on hierarchical and
multi-agent orchestration mechanisms, distributed mechanisms that build upon decentralized
intelligence, or federated orchestration involving interactions between multiple orchestrators
under different administrative domains.

Figure 5.10: Resource orchestration in the continuum [HEX224-D63]

In the past years Edge and Internet of Things (IoT) computing have arisen as a paradigm that
aims to provide computing, storage and networking capabilities in near proximity of the end-
users, while providing the same pay-as-you-go model of Cloud computing. While edge computing
enables application developers and content providers to leverage Cloud computing capabilities
and an IT service environment at the edge of the network, IoT computing distributes resources
and services across Cloud, Edge, and devices on the field to create the so-called compute
continuum. The Management and Orchestration concepts currently applied to networks must be
expanded in 6G to meet the requirements of the compute continuum.

Integration and orchestration of the extreme edge resources in the compute continuum
enables the M&O capabilities of computing resources beyond the radio access part of the network.
It provides the architectural components, interfaces, and mechanisms needed to orchestrate and
manage the volatile and resource constrained extreme edge devices as part of the computing
continuum. For example, in the Immersive Experience, Trusted Environments and Fully

Dissemination level: Public | 76

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Connected World Use Case Families this enabler contributes towards the improvement of privacy
and security protection. Different applications from the above-mentioned use cases will be
allowed to handle their sensitive information (e.g., eHealth) in the device where it has been
generated. In addition, in the Collaborative Robots and Digital Twins Use Case Families this use
case contributes towards improved M&O and service continuity capabilities by providing
mechanisms for flexible resource inclusion and allocation in the compute continuum. The
compute continuum will have deep implications in the 6G architecture where new interfaces and
mechanisms need to be defined for: 1) exposing the capabilities of the extreme edge devices and
2) for interactions between the extreme edge, edge, and cloud resources.

In the area of network automation, zero-touch closed loops (CL) implement the logic for self-
configuration, self-adaptation, and self-optimization, towards autonomous, scalable management
of dynamic and multi-technology networks. CL functions are deployed on-demand and interact
with other M&O functions (e.g., monitoring, data analytics, digital twin, AI/ML functions) to build
the four stages of automation workflows: Monitoring, Analysis, Decision, and Execution. CLs can
be specialized for several objectives, e.g., SLA or intent assurance, resource usage optimization,
etc. They can work in reactive, proactive or predictive mode and operate with different time
scales, applied to different layers or domains. CL coordination is fundamental to guarantee the
consistency and efficiency of decisions coming from concurrent and interdependent CLs, applying
techniques for conflict detection and mitigation, or arbitration strategies.

5.3.3 NATIVE AI - PERVASIVE MONITORING SYSTEM

An AI-native 6G system architecture is defined by intelligence everywhere, distributed data

infrastructure, zero touch management and AI as a Service [6GIA24-Vision]. This architecture
enables AI/ML capabilities throughout the network, from central nodes to edge devices,
supported by a robust data infrastructure for data availability, observability, pre-processing, and
model lifecycle management across network layers.

In the context of DESIRE6G, we focus on delivering an AI-native 6G system architecture,

revisiting the orchestration management, control and data planes (see Figure 5.11). Following
the definition and principles outlined in [ERICSSON23], to make the system perceptive, DESIRE6G
introduces a pervasive monitoring system that extends to the user equipment, leveraging in-band
network telemetry solutions enabled by data plane programmability for precise, end-to-end
information collection. Data access is provided at multiple layers with varying granularity to
support decision-making across different levels and timescales, enabling operations at scale. We
use an AI-driven Service Management and Orchestration layer (SMO) that supports non-RT
decision-making, optimizations and MLOps. We further employ the use of a use of Multi-Agent

Dissemination level: Public | 77

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Systems (MAS) to support intelligent near real-time control loops, pushing network decision
making closer to the data plane [BRV24]. This functional split promotes service assurance
through enabling faster control loops, ensuring the scalability of the system as its self-* operation
relies mainly on the autonomous coordinated operation of the agents.

At the SMO level, the Optimization Engine is the entity responsible for generic optimizations
on medium to long timescales (> 1 sec), while the ML Function Orchestrator (MLFO) is
responsible of deploying and, if needed, reconfiguring the MAS of a given service; it is in charge
of creating AI/ML pipelines and relating them to the target service. AI/ML pipelines are
associated to network entities and need to be deployed and reconfigured properly according to
needs, e.g., flow rerouting of a service requires moving agents (with their performance data and
models) among different DESIRE6G sites. Service assurance is achieved mainly by the service-
specific MAS which implements distributed network intelligence closer to the physical
infrastructure. MAS is responsible for receiving service-specific monitoring information and fine-
tuning the network and compute resources to meet service-level KPIs (e.g. routing [BSM+24],
elastic scaling of computing resources [HMP+23]). It configures and uses the pervasive telemetry
system to receive service performance indicators, e.g., end-to-end latency for latency-sensitive or
latency-critical services.

Additionally, we integrate edge AI capabilities to optimize network services (e.g., RIS

configuration [CSB23]) and applications. This includes extending the architectural framework to
support in-network machine learning (ML) through the Infrastructure Management Layer that
acts as a combination of a Virtualized Infrastructure Manager (VIM) and a hardware abstraction
(HAL) layer and the integration of frameworks like SOL and VACCEL, which facilitate rapid cross-
framework and cross-hardware execution of AI tasks. Through these innovations, DESIRE6G
promotes pervasive AI, fostering collaborative intelligence across the network infrastructure.
The description of the DESIRE6G architecture and respective architectural components is
provided in [DESIRE24-D22].

Dissemination level: Public | 78

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 5.11: AI-native architecture [DESIRE24-D22]

5.4 DIGITAL TWIN

This section discusses several methos on how to introduce Digital twin in the network. Digital
twins enable a virtual environment where e.g., AI-driven network functions can be developed,
tested, and optimized without jeopardizing the actual network performance. The DT can support
automation, real-time decision-making, and dynamic resource allocation through the integration
of Network Digital Twins.

5.4.1 INTEGRATING NETWORK DIGITAL TWINS INTO 6G

ARCHITECTURES

The increasing complexity and dynamic nature of communication networks as we move

towards 6G presents several challenges. Traditional network management methods are
becoming insufficient in handling the scale and variety of modern networks, particularly when
real-time, high-speed data processing is required.

The challenge also stems from the limitations in current network simulation tools, which are
typically designed for specific use cases or domains and often lack the scalability required for
larger and more complex scenarios. Current approaches do not fully account for the integration
of AI-based management or the seamless coupling between the physical and digital network
elements. As a result, networks struggle to adapt to rapidly changing environments, making it
essential to develop new architectures that can support automation, real-time decision-making,
and dynamic resource allocation through the integration of Network Digital Twins (NDTs)
[TCF24].

Dissemination level: Public | 79

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

The concept of NDTs, when integrated into a 6G architecture, addresses these challenges by
providing a virtual environment where AI-driven network functions can be developed, tested,
and optimized. This approach allows for real-time adaptation and automation, ensuring networks
can evolve and scale to meet the demands of future applications [Fay+24].

To address the outlined challenges, the initial proposed architecture introduces several key
components that extend existing standard development organization (SDO) architectures. One of
the main architectural innovations is the integration of a Network Digital Twin (NDT), which
operates across three distinct layers: the physical network, the digital network, and a federated
simulation framework.

The physical layer remains consistent with existing network elements, such as User
Equipment (UE), RAN, and core network, while the digital layer introduces a network twin that
allows for dynamic simulation and control. The digital layer is built upon the ITU-T Y.3090
recommendation [ITUT22-Y3090], which outlines two core model types: basic and functional
models. As defined in [ZST24]:

• A basic model of a network element is the collection of data describing its properties,
configurations, and operational status, along with any associated algorithms or protocols
used to emulate its dynamics and evolution with time. A basic model of a network is the
aggregation of basic models of network elements, including their physical and logical
relationships and the interactions that occur between them.

• A functional model of a network builds upon basic models, applying advanced

processing techniques, often through AI/ML algorithms, under varying operational
scenarios. These models are designed for specific objectives such as performance
optimization, anomaly detection, or predictive maintenance.

The third layer, the federated simulation framework, enables the coupling of multiple domain-
specific simulators, forming a unified system that allows for large-scale scenario testing. This
framework supports both online and offline NDTs, enabling networks to perform "what-if"
analyses and refine AI-based functions before deploying them in real-world environments. This
is critical for the orchestration of AI-driven services, providing a feedback loop for real-time
performance optimization.

Dissemination level: Public | 80

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 5.12: illustrates a high-level architecture of the proposed solution, highlighting the
interaction between various layers and the integration of NDTs for closed-loop management and
control

5.4.2 AI-DRIVEN AND MLOPS-ENABLED NETWORK DIGITAL TWIN FOR

MULTI-DOMAIN COMMUNICATIONS

To enhance deterministic communications, the combination of AI/ML capabilities with

reliability and time sensitiveness aspects can be introduced in the 6G system to simulate and
predict future network status, including KPI prediction for provisioning services, simulation of
different types of flows, user demands, network congestion, etc. These capabilities can also be
integrated with Network Digital Twins (NDT, [IRTF-Zhu+24]) and Monitoring modules, where
data can be generated or ingested for training and generating AI/ML models or inferences, or for
assessing and validating the outcomes of the models generated across the different network
domains. Indeed, the use of a Network Digital Twin (NDT) technology can help in estimating the
achievable performance of deterministic services traffic flows in near-real-time. The
implementation of a NDT in production environments faces several challenges and requirements.
Among others, it includes: i) high accuracy in the estimating/prediction of the performance of
both already deployed and requested traffic flows; such performance can be defined in terms of
Key Performance Indicators (KPI), such as end-to-end delay and delay variation; ii) the ability to

Dissemination level: Public | 81

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

easily define scenarios for KPI estimation; and, iii) short computation time, (e.g., under 1 min), to
provide KPIs when traffic conditions vary, for example, with time.

The introduction and serving of AI/ML capabilities and its combination with NDTs and
Monitoring components across multiple technology domains poses a new challenge towards the
design of 6G system architecture. Such combination of techniques allows to understand the
network infrastructure behaviour, anticipate future states and steer the network infrastructure
towards the desired goal in terms of service performance.

The concept of MLOps is defined in literature as the technology paradigm based on the
extension of the DevOps methodology for enabling the full lifecycle management of AI/ML models
in production environments. As part of this approach, MLOps targets the integration of different
processes related to data collection, data transformation or model training towards the
development and deployment of AI/ML models [EAD14].

Leveraging this technology concept and considering the significant role to be played by AI/ML
techniques in 6G systems [BKJ+23], the design and integration of a specific AI/ML framework
solution applying MLOps main principles is seen as a key enabler to support and facilitate the
development and delivery of AI/ML services across multiple 6G System domains.

The current contribution proposes the integration of an AI/ML framework in the Management
and Orchestration domain control loops of 6G architecture to enable the management of the
complete lifecycle of AI/ML models, from design and training to deployment and serving in
production environments at different domains. Serving of AI/ML capabilities can be shared in the
shape of service either through the serving of AI/ML models in the shape of artifacts or through
the delivery of AI/ML inference results, i.e., the resulting predictions or model outputs [BGG+23].

In the context of enhancing determinism, the introduction of AI/ML framework allows the
training and distribution of AI/ML models focused on the simulation and prediction of service
performance KPIs. The integration of these models with a Network Digital Twin architecture
component would allow to trigger the simulation of service KPIs before performing path
computation processes as part of the deterministic service provisioning stage. Additionally,
monitoring information flows can be used to evaluate and retrain AI/ML models according to the
real performance obtained once applied specific network and resource configurations for
provisioning deterministic services.

From the architectural point of view, the proper deployment of a NDT in support of service
KPIs prediction/estimation, requires the usage of standardized interfaces to assure: i)
interoperability between NDT and the underlying real network/system to enable real-data
monitoring and collection to allow, for example, NDT models adjustment, and ii) integration

Dissemination level: Public | 82

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

between NDT and the network Control Plane (e.g., Service Automation, Path Computation,
Monitoring, etc.) to provide support for the internal communications among control plane
architectural components to enable service provisioning workflows.

Figure 5.13: MLOps-based AI/ML framework in support of Digital Twinning

The design of the AI/ML framework solution is proposed as a unified AI/ML platform based
on a service-oriented design that allows to design, train, serve and monitor different types of
AI/ML algorithms (e.g. supervised, unsupervised, hybrid, distributed, etc.), in different network
domain as part of the Management & Orchestration domain control loops.

To offer these functionalities, AI/ML framework solution is composed of six main blocks:
Pipeline Development, Pipeline Orchestration Framework (POP), Model Storage, Model Serving,
Inference System, and Model and Metrics Monitoring. The outputs of the framework can be
served to any domain in the shape of models via the REST/gRPC interface available in Model
Storage module, or in the shape of inferences via the REST interface available in the Inference
System module [PREDICT23-D31], [PREDICT24-D32]. In the context of determinism, this
solution is integrated with the Network Digital Twin (NDT) and Monitoring frameworks, being
the NDT and the Monitoring in charge of feeding the models for obtaining Service KPI simulations
or predictions and for validating the results obtained.

5.5 REFERENCES
[3GPP24-23.501] 3GPP, “System Architecture for the 5G System (5GS)”, 3rd Generation Partnership
Project, Technical Specification TS 23.501, ver. 19.1.0, September 2024.
[6GIA24-Vision] Uusitalo, M., Bernardos, C. J., Kaloxylos, A., Bourse, D. A., Norp, T., Lønsethagen, H., Hecker,
A., Rugeland, P., Papagianni, C., Bulakci, Ö., Li, X., Ericson, M., Anton-Haro, C., Massod Khorsandi, B., Ramos-
Lopez, A., Frascolla, V., Marco, G., Gavras, A., & Trichias, K. (2024). European Vision for the 6G Network
Ecosystem. Zenodo. https://doi.org/10.5281/zenodo.14230482
[AHM24] R. Adeogun, H. Hrasnica, C. N. Manchon, The CENTRIC Project Vision on Sustainable AI-native Air-

Dissemination level: Public | 83

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Interface for 6G Networks, Feb 2024, aval: https://centric-sns.eu/wp-

content/uploads/2024/02/centric_first_vision_paper_v1.pdf
[ARF+24] Ilias Alexandropoulos, Vasiliki Rentoula, Dimitrios Fragkos, Nikolaos Gkatzios, Harilaos
Koumaras, “An AI-assisted User-Intent 6G System for Dynamic Throughput Provision”, IEEE International
Workshop on Computer-Aided Modeling, Analysis, and Design of Communication Links and Networks,
CAMAD 2024
[5BDG+14] Bosshart, P., Daly, D., Gibb, G., Izzard, M., McKeown, N., Rexford, J., ... & Walker, D. (2014). P4:
Programming protocol-independent packet processors. ACM SIGCOMM Computer Communication Review,
44(3), 87-95.
[BEG23-D41] BeGREEN Deliverable D4.1, “State-of-the-Art Review and Initial Definition of BeGREEN O-
RAN Intelligence Plane”, December 2023, [Online] Available: https://www.sns-
begreen.com/deliverables?id=1008107
[BEG24-D42] BeGREEN Deliverable D4.2, “Initial Evaluation of BeGREEN O-RAN Intelligence Plane, and
AI/ML Algorithms for NFV User-Plane And Edge Service Control Energy Efficiency Optimization”, [Online]
Available: https://www.sns-begreen.com/deliverables?id=1031615
[BGG+23] Bahare, M. K., Gavras, A., Gramaglia, M., Cosmas, J., Li, X., Bulakci, Ö., Rahman, A., Kostopoulos, A.,
Mesodiakaki, A., Tsolkas, D., Ericson, M., Boldi, M., Uusitalo, M., Ghoraishi, M., & Rugeland, P. (2023). The 6G
Architecture Landscape - European perspective. Zenodo. https://doi.org/10.5281/zenodo.7313232
[BKJ+23] Ankur Bang, Kapil Kant Kamal, Padmaja Joshi, Kavita Bhatia, 6G: The Next Giant Leap for AI and
ML,Procedia Computer Science, Volume 218, 2023, Pages 310-317, ISSN 1877-0509,
https://doi.org/10.1016/j.procs.2023.01.013.
[3BKL+24a] F. Brisch, A. J. Kassler, S. Laki, P. Hudoba, G. Pongrácz, P4-MTAGG - a Framework for Multi-
Tenant P4 Network Devices, Würzburg Workshop on 6G Networks (WueWoWAS'24), 30 September-1
October 2024, Würzburg, 2024.
[4BKL+24b] F. Brisch, A. J. Kassler, S. Laki, P. Hudoba, P4-MTAGG - a Framework for Multi-Tenant P4
Network Devices, 20th International Conference on Network and Service Management (CNSM 2024),
DEMO PAPER, 28-31 October, Prague, Czech Republic, 2024.
[BRV24] S. Barzegar, M. Ruiz, and L. Velasco, "Autonomous Flow Routing for Near Real-Time Quality of
Service Assurance," IEEE Transactions on Network and Service Management (TNSM), vol. 21, pp. 2504-
2514, 2024.
[BSM+24] S. Barzegar, H. Shakespear-Miles, F. Alhamed, F. Paolucci, P. Gonzalez, M. Ruiz, and L. Velasco,
"Near Real-Time Autonomous Multi-Flow Routing with Dynamic Optical Bypass Management," in Proc.
International Conference on Optical Network Design and Modeling (ONDM), 2024.
[CENTR24-D21] S. Cammerer and J. Hoydis, “D2.1 Evaluation of GPU Implementation of ML-based
Receiver”, Zenodo, Jul. 2024. doi: 10.5281/zenodo.12773161.
[CSB23] C. B. Chaaya, S. Samarakoon and M. Bennis, “Federated Learning Games for Reconfigurable
Intelligent Surfaces via Causal Representations”. In GLOBECOM 2023-2023 IEEE Global Communications
Conference, Kuala Lumpur, Malaysia, 2023, pp. 6567-6572.
[DESIRE24-D22] Pongrácz, G., & Papagianni, C. (2024). D2.2: DESIRE6G Functional Architecture Definition.
Zenodo. https://doi.org/10.5281/zenodo.12784579
[DESIRE24-D22] Pongrácz, G., & Papagianni, C. (2024). D2.2: DESIRE6G Functional Architecture Definition.
Zenodo. https://doi.org/10.5281/zenodo.12784579
[EAD14] Erich F., Amrit C., Daneva M., “A Mapping Study on Cooperation between Information System
Development and Operations”, In: Jedlitschka A., Kuvaja P., Kuhrmann M., Männistö T., Münch J.,
Raatikainen M. (eds) Product-Focused Software Process Improvement. PROFES 2014. Lecture Notes in
Computer Science, vol 8892. Springer, Cham, 2014.
[ERICSSON23] Ericsson. (2023). “Defining AI native: A key enabler for advanced intelligent telecom
networks [White paper]. https://www.ericsson.com/en/reports-and-papers
[Fay+24] S. Faye et al., "Integrating Network Digital Twinning into Future AI-based 6G Systems: The 6G-
TWIN Vision," 2024 Joint European Conference on Networks and Communications & 6G Summit

Dissemination level: Public | 84

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

(EuCNC/6G Summit), Antwerp, Belgium, 2024, pp. 883-888, doi:

10.1109/EuCNC/6GSummit60053.2024.10597058.
[GKF+24] N. Gkatzios, H. Koumaras, D. Fragkos and V. Koumaras, "A Proof of Concept Implementation of an
AI-assisted User-Centric 6G Network," 2024 Joint European Conference on Networks and Communications
& 6G Summit (EuCNC/6G Summit), Antwerp, Belgium, 2024, pp. 907-912, doi:
10.1109/EuCNC/6GSummit60053.2024.10597020.
[GNT+224] P. Gkonis, N. Nomikos, P. Trakadas, L. Sarakis, G. Xylouris, X. Masip- Bruin, and J. Martrat,
“Leveraging network data analytics function and machine learning for data collection, resource
optimization, security and privacy in 6g networks,” IEEE Access, pp. 1–1, 2024.
[HEX223-D33] HEXA-X-II Deliverable D3.3, “Initial analysis of architectural enablers and framework”
https://hexa-x-ii.eu/wp-content/uploads/2024/04/Hexa-X-II_D3.3_v1.0.pdf, April 30, 2024
[HEX224-D63] HEXA-X-II Deliverable D6.3 – Initial Design of 6G Smart Network Management Framework,
2024. Online. https://hexa-x-ii.eu/wp-content/uploads/2024/07/Hexa-X-II_D6-3_v1.0.pdf
[HMP+23] Hsu CS, Martín-Pérez J, Papagianni C, Grosso P. V2n service scaling with deep reinforcement
learning. InNOMS 2023-2023 IEEE/IFIP Network Operations and Management Symposium 2023 May 8
(pp. 1-5). IEEE.
[IRTF-Zhu+24] C. Zhou et al., “Network Digital Twin: Concepts and Reference Architecture”, draft-irtf-
nmrg-network-digital-twin-arch-06”, July 2024.
[ITUR19-NET2030] White Paper Network 2030: A Blueprint of Technology, Applications and Market
Drivers Towards the Year 2030 and Beyond. FG-NET-2030 group - ITU-R, 2019. Available online at
https://www.itu.int/en/ITU-T/focusgroups/net2030/Documents/White_Paper.pdf
[ITUR23-M2160] ITU-R, “Framework and overall objectives of the future development of IMT for 2030 and
beyond”, M.2160, November 2023. [Online]. Available: https://www.itu.int/rec/R-REC-M.2160/en.
[ITUT22-Y3090] ITUT, “Digital twin network – Requirements and architecture.,” ITU, recommendation
ITU-T Y.3090, 2022.
[KCC+22] T. Kellermann, R. P. Centelles, D. Camps-Mur, R. Ferrús, M. Guadalupi and A. C. Augé, "Novel
Architecture for Cellular IoT in Future Non-Terrestrial Networks: Store and Forward Adaptations for
Enabling Discontinuous Feeder Link Operation," in IEEE Access, vol. 10, pp. 68922-68936, 2022, doi:
10.1109/ACCESS.2022.3184720.
[LST+20] [2] T. Li, A. K. Sahu, A. Talwalkar and V. Smith, “Federated Learning: Challenges,
Methods, and Future Directions,“ IEEE Signal Processing Magazine, vol. 37, no. 3, pp. 50-60, May 2020, doi:
10.1109/MSP.2020.2975749.
[Mao+22] B. Mao et al., “AI Models for Green Communications Towards 6G,” IEEE Communications Surveys
Tutorials, vol. 24, no. 1, 2022.
[Nt+24] K. Ntontin et al., "ETHER: A 6G Architectural Framework for 3D Multi-Layered Networks," 2024
IEEE Wireless Communications and Networking Conference (WCNC), Dubai, United Arab Emirates, 2024,
pp. 1-6, doi: 10.1109/WCNC57260.2024.10570731
[ORAN23-nGRG] O-RAN Alliance, “Research Report on Native and Cross-domain AI: State of the art and
future outlook”, nGRG group, September 2023.
[2P4R24] P4Runtime Specification. 2024. https://p4.org/p4-spec/p4runtime/main/P4Runtime-
Spec.html
[PREDICT23-D31] PREDICT-6G Consortium. (2023). D3.1 Release 1 of AI-driven inter-domain network
control, management, and orchestration innovations. Zenodo. https://doi.org/10.5281/zenodo.12167712
[PREDICT24-D32] PREDICT-6G Consortium. (2024). D3.2 Implementation of selected release 1 AI-driven
inter-domain network control, management and orchestration innovations. Zenodo.
https://doi.org/10.5281/zenodo.12167665
[SAFE6G] Project SAFE-6G. Online. https://safe-6g.eu/
[STK+24] M. M. Saad, M. A. Tariq, M. T. R. Khan and D. Kim, "Non-Terrestrial Networks: An Overview of
3GPP Release 17 & 18," in IEEE Internet of Things Magazine, vol. 7, no. 1, pp. 20-26, January 2024, doi:

Dissemination level: Public | 85

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

10.1109/IOTM.001.2300154.
[Str+24] E. C. Strinati et al., "Goal-Oriented and Semantic Communication in 6G AI-Native Networks: The
6G-GOALS Approach," 2024 Joint European Conference on Networks and Communications & 6G Summit
(EuCNC/6G Summit), Antwerp, Belgium, 2024, pp. 1-6, doi:
10.1109/EuCNC/6GSummit60053.2024.10597087.
[TCF24] I. Turcanu, G. Castignani and S. Faye, "On the Integration of Digital Twin Networks into City Digital
Twins: Benefits and Challenges," 2024 IEEE 21st Consumer Communications & Networking Conference
(CCNC), Las Vegas, NV, USA, 2024, pp. 752-758, doi: 10.1109/CCNC51664.2024.10454704.
[ZST24] A. Zaki-Hindi, JS. Sottet, I. Turcanu, and S. Faye, " Network Digital Twins for 6G: Defining Data
Taxonomy and Data Models," 2024 IEEE Conference on Standards for Communications and Networking
(CSCN), Belgrade, Serbia, 2024.

Dissemination level: Public | 86

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

6 INTEGRATED SENSING AND

COMMUNICATION
6G services will be associated with a wide spectrum of vertical applications with greatly
varying requirements and will offer advanced features beyond connectivity spanning from
sensing to monitoring and positioning. To address these requirements, 6G will feature Integrated
Sensing and Communication (ISAC) capabilities, performing sensing through the mobile
communication infrastructure. This can be ac achieved adopting either a Channel State
Information (CSI) or a passive radar approach. The CSI sensing approach relies on the
connectivity established between the base station (BS) and the user equipment (UE) to estimate
channel conditions and extract information for the Angle of Arrival (AoA) and the Time Difference
of Arrival (TDoA). This information can be then used to support a set of applications1 including
human localization and tracking, presence detection, activity recognition, healthcare, etc. In the
“radar” sensor approach, the network exploits its own radio signals to sense and comprehend the
surrounding physical world. The echoes (reflections) and scattering of wireless signals
predominately transmitted for communication purposes, provide information related to the
characteristics of the environment and/or objects therein [3GPP24-22.837]. The sensing data
collected and processed by the network can then be leveraged to enhance the operations of the
network, augment existing services such as XR and digital twinning, and enable new services,
such as object detection and tracking, along with imaging and environment reconstruction. This
potential has already attracted a lot of attention from 3GPP, which has initiated a preliminary
study on use cases and ISAC requirements, making it a promising candidate to optimize both
communications and sensing systems [3GPP24-22.837].

Depending on the level of integration of the sensing functionality into the communication
network, different approaches can be adopted including:

• fully separated infrastructures performing sensing and communications functionalities.

Based on this approach, information acquired from one infrastructure is used to assist the
other.

• common hardware supporting sensing and communication capabilities. This approach is

implemented by sharing the available spectrum, with the constraint that sensing, and
communication signals are transmitted over different timeslots.

1 An Overview on IEEE 802.11bf: WLAN Sensing

Dissemination level: Public | 87

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

• fully integrated systems sharing both spectrum and time domains.

Depending on the number and roles of the devices involved in sensing several options also
exist including:

• the monostatic case where a single device is used for transmitting and receiving sensing
signals

• the bi-static/multi-static sensing where a single transmitter and physically separated

single or multiple receivers are used to acquire the sensing signals.

• the passive sensing approach where signals transmitted primarily for communication
purposes can be also used by other devices for sensing.

6.1 INTEGRATION OF NON-3GPP AND 3GPP SENSING

Although some early prototypes are available validating these concepts, these are mostly
available in non-3GPP networks (i.e., Wi-Fi), whereas implementations of 3GPP compliant passive
radar-based ISAC systems are still at a very early stage. The main reason is that these systems
demand additional complexity in signal processing and require collection and aggregation of huge
volumes of synchronized in-phase and quadrature (IQ) reflected (echo) streams that need to be
processed to extract information on the sensed environment. This processing can only be
performed at edge servers, introducing the need to transport the IQ streams over flexible high-
capacity transport networks.

rAPPs
SMO rAPPs
Core to
SMO
(O1-extension)

A1 Y1
RIS to SMO

A1
O1

Transport Near-RT RIC Transport Transport

Netconf Agent Real time RIC Core Network
Control Control Control
DBaaS
CF-MIMO xapp xapp
Sensing Comms Monitoring
E2

E2

RIS RU E2
E2

E2
E2

DU
O-CU
Transport Unit
Transport Unit
Transport Unit
Transport Unit

Transport Unit

Transport Unit
Transport Unit

Transport Unit
Transport Unit

RU Comms (O-FH) F1 F1 N2
DU
Obstacle

RU O-DU DB

UE
Sensing I/Q sensing
RU
sensing
3GPP
UPF DN

E2
WiFi N3IWF Transport network
Transport network Transport network Core
Non-3GPP
O-cloud

Figure 6.1: Generic architecture of integration of 3GPP and non-3GPP in the core network
[6GSENSESD2.2]

Figure 6.1 proposes a 6G architecture that interconnects a multi-technology Radio Access

Network (RAN) able to offer sensing functionalities (3GPP and non-3GPP) with core network
domains, to facilitate joint support of sensing and communication services. The RAN technologies
of interest include non-3GPP (Sub-6, Wi-Fi, mmWave) and 3GPP based (5G NR) networks, which

Dissemination level: Public | 88

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

will coexist in an ISAC framework with the to obtain accurate representation of the surrounding
environment.

Non-3GPP based sensing is performed adopting Wi-Fi networks which are extended to operate
as monostatic and multi-static radars. The output of the sensing information from non-3GPP
networks is exposed to the RAN segment in a secure way through suitable extensions of the E2
interface of the RAN Intelligent Controller (RIC). To achieve this, there is a need to enable Wi-Fi
networks to expose their sensing related data in a secure way. Addressing this requirement, in
[6GSENSESD2.2] it has been proposed the use of the non-3GPP Inter-Working Function (N3IWF),
which is responsible for interworking between untrusted non-3GPP networks and the 5G core.
6G-SENSES, therefore, proposes to adopt and appropriately extend N3IWF, in order to provide
the necessary access and authentication protocols with new features that will allow Wi-Fi networks
to securely expose sensing data to the RIC. 3GPP-based sensing is performed based on the principle
of a distributed passive wireless radar. According to this, 6G BSs generate communication signals
reflected on “objects” located in the surrounding area, creating IQ echo streams. These IQ echo
streams are transmitted in the form of uplink fronthaul streams to the DUs, where they are
compressed (down-sampled) and transmitted through the E2 interface to the RIC. Purposely
developed sensing xApps fuse the incoming sensing streams (IQ echo streams and Wi-Fi sending
data), analyse their quality and cache data to a fast in-memory database. These data can be then
exploited internally by the system to optimize the operational parameters of the various building
blocks of the RAN segment (e.g. development of beamforming, beam steering, power control, etc.),
or they can be exposed to the vertical applications through the Y1 interface.

The sensing output is also passed to the Service Management and Orchestration (SMO), that
decides the optimal network resource configuration to support both communication and sensing
services. To perform this, the SMO provides mechanisms supporting automated lifecycle
management (LCM) for ISAC services instantiating and automatically reconfiguring E2E slices
considering both communication (i.e. fronthaul backhaul) and sensing services requirements. A
fist concept demonstration of this architectural approach is detailed in [AGT25].

6.2 DISTRIBUTED SENSING ARCHITECTURE

The goal of a distributed sensing system is to perform distributed and cooperative sensing,
involving collection and exploitation of data sensed by multiple heterogeneous devices, i.e.,
sensing receiver nodes (SRNs), as well as tracking many heterogeneous (and mobile) targets,
including both passive and active user equipment (UEs), over a large area. In this regard, an
important focus is to address the challenge of a lack of suitable sensing control functions for
distributed operation.

Dissemination level: Public | 89

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Sensing data is initially collected at the Transmission Reception Point (TRP), where some local
processing occurs before measurement data is transmitted to the Sensing Management Function
(SeMF) in the core network. The two main approaches include i) Information-Level Fusion,
involving preliminary detection and estimation at each local station (e.g., Angle of Arrival and
Time of Arrival), and transmission to a fusion centre (FC). The FC performs global detection and
localization through pairing and triangulation of the extracted parameters, and ii) Signal-Level
Fusion, involving directly and jointly processing the local raw observations at the FC for target
detection and localization, thus avoiding the need for local decision fusion or target-measurement
association [Y+23].

The trade-off between sensing performance and the volume of data to be stored and
exchanged represents a major challenge when choosing the approach to be used based on the
requirements of the application. Moreover, mobility of passive targets, i.e., objects that are not
connected with the network, brings the challenge of handover while tracking the objects over
space and time to ensure continuity of sensing service.

Another key challenge to be explored is represented by including semantic and goal-oriented

functionalities. This impacts the architecture design, requiring the inclusion of a semantic plane,
which plays a pivotal role by managing the extraction, interpretation, and transmission of
contextual meaning, thereby ensuring that the transmitted information is meaningful and aligned
with the system's objectives, rather than just raw data. Indeed, the challenge is also that to include
effectively the semantic plane into a distributed integrated communication and sensing
architecture, interfaces and new semantic modules are required to ensure that sensing,
communication, and computation tasks are aligned with the system's semantic goals.
Additionally, the semantic plane handles the dynamic adaptation of meaning in changing
environments, maintaining coherence across distributed nodes, and optimizing the network's
overall performance.

The design of such distributed sensing methods involves accommodating for heterogeneous
computing capacities, energy budget constraints, as well as sensing and communication
capabilities. In this regard, a key example is given by the integration of different types of
reconfigurable intelligent surfaces (RIS) [CSA+21] (e.g., passive, active, hybrid, or autonomous
[ARD+22]). While existing works generally focus separately either on the communication
optimisation or the sensing accuracy, or on spectrum sharing techniques, the challenge is to
develop novel methods to enable joint optimization of ISAC, while accounting for such
heterogeneous characteristics.

Dissemination level: Public | 90

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 6.2: Distributed ISAC Architecture [C+25]

Figure 6.2 shows the major features of the distribute architecture – distributed and intelligent
processing, the use of the semantic plane for control and the role of RIS.

The distributed processing involves collaboration among intelligent nodes to provide a multi-
perspective and integrated view of the environment of interest, including both active UEs and
passive objects, extracting both channel-related propagation parameters, such as angles, delays,
Dopplers, and geometric information like location, orientation, and size of the targets. This entails
not just fusion of data from multiple SRNs, but also a map of the environment. Given the high
dimensionality of the sensing data on MIMO systems, the data transformation is performed at the
local (LDP) level to ensure that the communication capacity of the system is not impacted. At the
same time, the architecture is flexible enough to support low-capacity nodes, which have limited
computing power, thus handling heterogeneous nodes and their associated hardware-related
limitations.

Devices that have different communication, sensing, computation, and storage capabilities as
well as power consumption are considered in the design of resource allocation and orchestration
schemes. This includes both energy-autonomous, energy-supplied, or energy-neutral STNs,
devices equipped with advanced or basic sensing capabilities (e.g., almost passive and reflective,
simultaneously reflecting and sensing, amplifying reflective RISs, massive MIMO). In the latter
case, the FC performs the data processing for those nodes. Furthermore, interfaces and protocols

Dissemination level: Public | 91

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

enable this intelligent sharing and cooperation among the different network elements. Such
architecture facilitates sharing of appropriate information from sensing modules to enable
substantially enhanced communications via joint optimization of both multi-antenna
transmissions and receptions, as well as the reflective beamforming of multi-functional RISs for
both localisation, sensing, and sensing-aided communications.

The semantic framework provides semantic processing to reduce these data volumes
intelligently over time and space. In the context of multi-modal sensing using heterogeneous
nodes, the framework provides semantic reasoning to efficiently integrate non 3GPP devices.

Using AI based reasoning, the semantic framework enables extraction of semantic information
that uses not only the sensor data but also previously obtained background knowledge (for
instance as an inference ML model trained from previous data and subject to rules imposed to the
observation environment). This will enable the adaptation of the ISAC sensing parameters
(refresh rate, performance criteria) and resource allocation to varying KPI/KVI requirements, as
well as selective information sharing (semantic information, extracted and processed given
accumulated background knowledge) and reasoning about multi-modal sensed information (i.e.,
generated by different sensor types).

6.3 OPTICAL WIRELESS COMMUNICATION BASED

ISAC
Localisation is one of the most essential sensing functions required for enabling the
Communications-Computing continuum in Industry 4.0 use cases such as automated guided
vehicles (AGV) Navigation for enhancing industrial automation process. In such scenarios, Optical
Wireless Communication (OWC) technology can be used as a supplementing radio technology
along with a sub cm location measurement and sensing solution to enable sensing and localisation
in the industry environment.

OWC ITU G.9991 networks are inherently cell-free with access to a cluster of 6 OWC photonic
antennas controlled by a single OWC access point to typically support a 4x5 meters coverage area.
This paves the way towards a beam-steered OWC system, which will ultimately produce much
wider access angles, which would subsequently require much fewer OWC APs. Increasing the
radius by a factor of 4 with a field of emission angle of 70°, which provides a coverage area from
one access point of 16x20m = 320m2 which would require 340 OWC AP clusters for a factory with
a floor space area of 109,100m2 which is much more commercially viable.

The main technique of localization with OWC is mainly the use of the received signal strength
(RSS), the TOA or time or phase difference of arrival (TDoA/PDoA) and the AOA to extract

Dissemination level: Public | 92

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

position information, either by trilateration or by triangulation. The RSS method is the most
popular and algorithms coupled with Kalman filters have been proposed to improve the
positioning resolution and responsiveness to motion. Furthermore, precise round trip (RT) TOA
measurements are possible, which are based on several APs and can be implemented using
existing fields in the physical layer preamble, together with fine timing measurement (FTM)
available in the MAC layer of IEEE Std. 802.11 (Wi-Fi). The aim is to achieve sub-centimetre
accuracy in 3D including the information of orientation angle of the target, based on several
improvements [OPTI24-D21]: a) develop a combined method with RSS and RT TOA, which, to our
knowledge, has not been addressed yet; b) combine this first sensing approach with AOA
measurements by modulating the angle of emission of the source and, thus, simplify the receiver
which is an important point in the case of simple embedded receivers.

Figure 6.3: OWC-based Localization and Sensing

Localization of UEs using ToA from sub-6 GHz wireless and RSS from Optical Wireless Infrared
techniques can be used to obtain an accuracy of less than 1 cm, however it is highly dependent on
the continuous direct line of sight access between the gNB access points and the UE. In [CAM+24]
the authors show in a laboratory experiment that 2 to 3 mm localization accuracy can be achieved
using ToA in a sub 6 GHz 5G Orthogonal Frequency Division Multiplexing (OFDM) Network. If
there is no direct line of sight access to four or more gNB access points then location ambiguity is
introduced, and so other techniques can be used to maintain localization such as AoA from
received radio signatures. It has been shown that a 2 - 3 degrees orientation accuracy can in
principle be obtained in a simulation experiment using Matlab’s Siteviewer 3D radio propagation
model [CAM+24]. If the received radio signatures estimate AoA from more than two access points
or AoA and distance from one access point, then some form of interim measure for obtaining
location can continue to be made. In [OPTI24-D41] the authors propose to test if more accurate
localization of UEs can be obtained using ToA from near IR OWC because of the greater system

Dissemination level: Public | 93

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

bandwidth that can be used to lower the levels of multipath reflections and noise which are
expected.

6.4 SENSING SERVICE PROVISIONING AND EXPOSURE

There is a recent interest into using the mobile network resources for services beyond
conventional connectivity. One example of this is the concept of integrated communication and
sensing (ICAS). The main challenges identified in the above scope are mostly related to the
capability of the network to support such challenging services in (i) a sustainable manner, and
(ii) a trustworthy, privacy-preserving manner, as large amounts of -beyond communications-
data will be required to be transmitted and processed, adhering to strict latency requirements. In
detail, the challenges include (a) the need for novel interfaces that support data collection, (b)
data processing, (c) data distribution & scaling of interfaces, (d) trust differentiation when
exposing to 3rd party applications, (e) network overload on the exposed Application
Programming Interfaces (APIs), (f) privacy risks, and (g) latency/performance challenges.

Beyond Communications Network (BCN) will enable new 6G services such as sensing and
compute offloading, and how to expose resulting data and relevant service capabilities in a secure,
privacy-preserving and efficient manner. The exposure and data management enabler therefore
aims to reduce the overhead from data exposure by aggregating and fusing data while ensuring
data privacy and trust. This enabler supports the creation of novel services that contribute to
societal benefits like safety and sustainability, supported by its capability to efficiently handle and
expose data from various producers, including the RAN and sensing nodes.

Any network entity with proper access rights should be able to access data or model(s), stored
from another entity. A form of authorization and/or authentication should be performed when a
network entity is trying to access, update or share data, analytics and model from another entity.
Security should be enabled E2E for any operation of the data collection services, including access,
exposure, storage, cleaning, processing and encoding. The network should be able to identify
energy-aware data collection services and facilitate their operations. Data collection and
exposure can be based on a local configuration, or a configuration received from the requester.
Different data consumers exist in the network (defined as general network entity), such as UEs,
RAN nodes, CN NFs, AFs, 3rd party applications, OAM, etc. Discovery, configuration and in some
cases evaluations of such data sources are among the functionalities to cover in 6G.

Further on, there is a need to develop novel APIs, enabling both internal network functions, as
well as third-party applications to request, receive and manage data securely and efficiently,
thereby reducing data traffic and overhead significantly. in the ISAC use case, exact information
on the position of base stations and UEs can be used to enable some kind of QoS-based sensing.

Dissemination level: Public | 94

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

However, when the positioning accuracy of measurements nodes is provided with low confidence
score, e.g., a UE position with some uncertainty, sensing services may have to be provided solely
by the network on a best-effort basis.

One way to improve sensing quality is obviously to carry out more radio measurements prior
to exposure of the measurement report to the requesting application, preferably measurements
that are geographically distributed. Involving more network nodes (UEs or base stations)
naturally leads to architectural challenges of centralized vs. distributed inference and processing
of the measurements.

The provision of sensing services by next generation communication systems necessitates the
introduction of a Sensing Management Function (SeMF) [HEXAD3.3], see Figure 6.4, that will be
responsible for facilitating an efficient coordination of sensing procedures, considering various
aspects such as sensing requirements, sensing capabilities, sensing constraints, etc. The SeMF can
be designed as a dedicated NF, since it is enabling a new functionality for next generation
networks. An alternative option would be to integrate the SeMF services as part of the location
management function of 5G. Since the amount of data from sensing may not be user specific, one
idea can be to transmit sensing data via a “data plane” that is routed via sensing functions in the
core network (and not via the UPF as for the user plane).

Figure 6.4: Functional Architecture, network-based sensing with UE involvement

At the same time, the need for identifying and coordinating compute workloads, in the context
of sensing service, AI or other computationally heavy tasks, in an efficient manner becomes
critical.

Different ISAC applications are associated with stringent requirements. On the one hand,
heavy computations need to be performed on the sensed data, coming from different sources, to

Dissemination level: Public | 95

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

provide with the information on their localization. On the other hand, ISAC applications could also
be associated with delay-strict requirements, where the results are expected to be received in
real-time (e.g., stopping a robot machine after detecting a human). In addition, although a far edge
cloud is located near the end user and comes with promise of reduced communication delay, it
can suffer from scarcity of compute resources which induces high processing delay. The trade-off
between network metrics and compute metrics would call for a new approach that enables the
Integration of Network and Compute (INC) to perform coordinated optimization.

The introduction of compute offloading in the next generation of networks should not increase
the complexity of the communication protocol. This can be achieved by tight integration and true
convergence of communication and computing, by introducing novel architectural components
for distributed computing. To satisfy the strict requirements on the computation and
communication latency, trustworthiness, power consumption and data accuracy, it is important
to introduce a common classification of computing and communication resources of each novel
component, as well as a common characterization of offloaded compute workload based on
predetermined requirements.

Figure 6.5: Integration of Network and Compute server collecting network and compute metrics
to decide optimised placement of sensing/other BC service consumer placement

When a device, acting as an offloading node, decides to offload a computation, it will have to
discover and select the candidate computation nodes, capable of performing the requested
computation while satisfying specific KPIs. Each computational node should estimate the task(s)
execution complexity and resources demand (i.e., computation and storage) based on a common

Dissemination level: Public | 96

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

characterization of the offloaded compute workloads (i.e., compute tasks) determined by the
requirements comprising traffic class (e.g., one time vs multi-iteration, one node vs multi-node,
etc.), computation complexity, communication requirements, precision requirement (e.g.,
quantization level of the compute data and operations), quality of compute service classes
(latency-sensitive, precision-sensitive, etc.).

The trade-off between network metrics and compute metrics would call for a new approach
that enables the Integration of Network and Compute to perform coordinated optimization.

6.5 ISAC APPLICATION FOR V2X USE CASES

The Euro NCAP announced that starting in 2024, new cars in Europe must include Vehicle-to-
Everything (V2X) connectivity for a five-star safety rating. In the U.S., the Department of
Transportation introduced a National V2X Deployment Plan in August 2024 to enhance road
safety, urging collaboration across industries and government. Despite European efforts to
deploy V2X, including contributions from major manufacturers, challenges like incompatible
technologies and limited consumer awareness have delayed progress. These challenges are now
being addressed, paving the way for broader V2X deployment.

The aim is to improve connectivity and safety for Fully Autonomous Vehicles (FAVs), especially
in city intersections, with goals of emergency vehicle route prioritization and optimized city-wide
traffic flow. The project requires real-time data sharing among road users and infrastructure,
dynamic traffic management, and the development of a high-capacity, ultra-low latency 6G-V2X
network. Key components include advanced LiDARs and RaDARs for object detection, Optical
Wireless Communication (OWC) for high-speed data exchange, and AI-based 3D mapping for real-
time environmental awareness.

Challenges for implementing technologies like LTE-V2X and OWC include achieving ultra-low
latency in dense environments, maintaining signal clarity in adverse weather, and integrating
diverse systems for seamless communication. Efficient computational resources and data
management will be crucial for real-time collective perception among vehicles and infrastructure.

Ultimately, the 6G architecture envisions a robust, AI-driven network supporting V2X in dense
urban areas, enabling collision-free navigation, improved traffic flow, and enhanced safety for
vehicles, Vulnerable Road Users (VRUs), and emergency services across interconnected city
areas.

Specifically, the focus on the support the next generation of (fully) autonomous driving aims
at the significant improvement of the road safety (and connectivity) for all road participants,
especially at busy city intersections. To achieve this, real-time, ultra-high speed and low latency

Dissemination level: Public | 97

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

with ultra-high reliability communication and data exchange between the FAVs, the VRUs and the
road/network infrastructure is mandatory. On top of the above, the system shall allow for
seamless coordination in critical scenarios like emergency vehicle prioritization and large-scale
vehicular traffic flow optimization across urban areas.

The proposed high-level network architecture, depicted in Figure 6.6, ensures that all road
participants (FAVs L0-L5, VRUs, etc.) can communicate between each other and with the
road/mobile network infrastructure, via either/both LTE/NR-V2X technologies (via Uu and/or
PC5) and/or via OWC/VLC, while the stringent QoS requirements posed by the safety related
applications -based on the collective perception concept- can be fulfilled by:

• The deployment of the OWC/VLC APs; throughput and latency-related.

• High-speed fronthaul/backhaul, photonic-based network which will ensure the resilient,

high-capacity, low-latency communication across the system; throughput, latency and
reliability-related.

• The co-deployment of LTE-V2X 5G/NR-V2X base stations/RUs and OWC/VLC APs and the
capability of the FAVs’ OBUs to transmit data concurrently via both radio technologies;
throughput, latency and reliability-related.

• Deployment of AI-based data processing infrastructure / data centres (incl. compute

nodes, application servers) capable of processing huge volumes of sensor data in real-
time, close to the road participants (see extreme/far-edge DCs); ultra-low latency
required for critical tasks such as, those related to JCAS and SLAM, accurate, real-time 3D
maps of the surrounding environment, decision-making for FAVs, real-time updates for
VRUs.

On top of the above, regional DCs and Central Cloud(s) are deployed at higher-levels, for data
aggregation and data processing, to support less critical applications such as, traffic flow
optimization at wider-city areas, but also utilized for traffic management, large-scale data
analytics, long-term storage, etc. The central cloud, in addition to the 4G/5G/6G core network
functions, it hosts the TMS which, by gathering and processing data from various sources, can
adjust the traffic lights and signals, aiming at congestion minimization and vehicular traffic flow
improvement along with emergency vehicle prioritization related functionalities. Finally, the AI-
based network management and orchestration functions will ensure the optimal allocation of the
optical network resources in real-time (and/or proactively) based on the current and predicted
traffic demands -to guarantee the applications’ QoS end-to-end, but also to improve the overall
network efficiency.

Dissemination level: Public | 98

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 6.6: ISAC-based V2X network architecture

In such deployment, fibre is essential as a high-bandwidth, low-latency medium connecting

Roadside Units (RSUs) to cloud processing platforms, enabling robust V2X communication. The
system relies on flexible optical networks, which support long-distance, low-latency data
transmission through optical amplifiers across a wide spectral band. However, fibre access is
limited near RSUs, so Free-Space Optics (FSO) is used to extend fibre-like connectivity to the last
meter. The architecture thus benefits from FSO’s ability to provide high-speed, fibre-equivalent
wireless communication for challenging network segments and enhances V2X services.

To bridge network gaps, a Fi-Wi-Fi bridge is used, allowing single-mode transmission through
transparent FSO channels, maintaining low latency, energy efficiency, and compatibility with
wideband optical transmission schemes. This connectivity is critical for high-demand
applications, such as 5G and 6G services at remote radio heads, by digitizing radio signals over
fibre or air.

In areas requiring longer-range communication, FSO is enhanced with MIMO or hybrid RF

links to ensure resilience against adverse weather. Short-range FSO applications connect RSUs in
dense urban areas, supporting high data loads without interfering with RF channels. Optical
Wireless Communication (OWC) complements RF by handling excess data traffic at intersections,
where high user density can cause RF congestion.

Fronthaul connectivity relies on quasi-coherent transceivers for high-speed, low-latency fibre

links, enabling real-time collective perception and SLAM processing in autonomous vehicles.
Initial deployment includes 50 Gb/s transceivers, with future upgrades to 100 Gb/s to handle
increased sensor data. These transceivers enable modular Bandwidth Variable Transceivers

Dissemination level: Public | 99

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

(BVTs) at central offices, delivering variable capacities of 2 Tb/s to 380 Tb/s across cell sites,
adapting to wireless network demands and future-proofing for data-heavy autonomous driving
applications.

6.6 MULTIX -ADVANCING ISAC THROUGH MULTI-

TECHNOLOGY, MULTI-SENSOR FUSION, MULTI-
BAND AND MULTI-STATIC PERCEPTION
The vision of 6G is to integrate sensing with communication in a single system. Sensing and
native Artificial Intelligence (AI) operations are the two key aspects to build the connected
intelligence in 6G. For sensing, the use of multiband, ranging from sub-6 GHz band and higher
frequency bands – from millimetre wave (mmWave) up to THz, wider bandwidth, and massive
antenna arrays will enable high accuracy and high-resolution sensing, which can help implement
the ISAC in a single system for mutual benefit. To enable such vision, the ability to utilize a range
of technologies to perform sensing and the ability to process the sensing data into meaningful
sensing results is a key challenge for 6G systems. Furthermore, how to manage such
heterogeneity and fully integrate various sensing sources and ISAC technologies throughout the
RAN up to the User Equipment (UE) across different layers of the RAN stack and thus explore the
full potential of each sensor, each node, each band, each technology as well as their combination
to achieve the maximum spectrum and system efficiency, energy efficiency and resource usage
across the entire RAN system.

To address the outlined challenges, a redesign of the 3GPP RAN system is required that can
aggregate a set of diverse design concepts in a seamless way (defining the so-called MultiX
concept) to create an integrated multi-sensor, multi-band, multi-static, and multi-technology
paradigm to enable multi-sensorial perception for future 6G sensing applications. Based on a
system architecture built on the upcoming 3GPP Release 20 (R20) and O-RAN specifications,
architectural design and develop enhancements are required to support aggregation of
perception data from multi-technologies and multi-sensors, their processing and exposure to
third parties in a secure, privacy-preserving and trustworthy way. Furthermore, support for
vertical and horizontal handovers and network selection procedures considering perception
requirements are envisaged, while exploring the use of AI / Machine Learning (ML) for novel
connectivity options to enhance the perception capabilities of the network.

Dissemination level: Public | 100

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Non-3GPP AN O-RAN

Network,
MPS Controller
UE Transport

Network,
Sensing PHY
Transport, App Data Access and
Security Hub (DASH)
High-Level

Sensign Plane
Sensing
Sensing Fusion

PHY 3GPP RU 3GPP DU 3GPP CU

Sensing Data
Processing
AI Engine RRC / SDAP
Sensing Results
Sensing Exposure
PDCP
Measurements

Figure 6.7: High-level architectural on 6G system design to enable multi-perception sensing

Supporting the above functionalities, data fabric and pipeline solutions are to be developed to
integrate mechanisms and artifacts to enable advanced data collection via novel paradigms (e.g.,
ProSE and Integrated Access and Backhaul (IAB)), processing, distribution, management (e.g.,
data curation and enrichment), and exposure for third-party access applications, ensuring
privacy, security, and trustworthiness. A Multi-Perception System (MPS) is envisaged to
providing perception to the network across different access technologies. The architectural
components such MPS will impact is illustrated in Figure 6.7 (MPS components in orange).
Current sensing approaches will be enhanced in the MPS-enhanced 6G system with multi-band,
OTFS-based, programmable environments and disaggregated architectures. Integration of multi-
technology RATs in the MPS, considering multi-static deployments, where synchronization is key.
AI is considered for the distributed processing of the raw data into actionable information,
including natively energy-efficient AI design of receiver architectures that jointly leverage on
signal processing domain knowledge and brain-inspired, event-based operation. All sensing data
is fused, processed and exposed in the Data Access and Security Hub (DASH).

6.7 REFERENCES
[3GPP24-22.837] V19.4.0 (2024-06) Feasibility Study on Integrated Sensing and Communication (Release
19).
[6GSENSESD2.2] 6G-SENSES Deliverable D2.2, “System architecture and preliminary evaluations”, March
2025.
[AGT25] Markos Anastasopoulos, Jesús Gutiérrez, Anna Tzanakaki, “Optical Transport Network
Optimization Supporting Integrated Sensing and Communication Services”, OFC 2025.
[Y+23] Yao, Shanliang et al. “Exploring Radar Data Representations in Autonomous Driving: A
Comprehensive Review.” (2023).
[CAM+24] John Cosmas, Kareem Ali, Hussein Malki, Adam Kapovits, Israel Koffman, Benjamin Azoulay,
Clément Lartigue, Emmanuel Plascencia, Bastien Béchadergue, Barthélemy Cagneau, Luc Chassagne,
Anastasius Gavras “Optical 6G Cell Free Networks for High Performance Communications and Sensing in
Industry 4.0” IEEE Future Networks World Forum 2024, 15-17 Oct 2024, Dubai UAE

Dissemination level: Public | 101

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

[CSA+21] E. Calvanese Strinati, G. C. Alexandropoulos, H. Wymeersch, B. Denis, V. Sciancalepore, R. D’Errico,

A. Clemente, D.-T. Phan-Huy, E. De Carvalho, and P. Popovski, “Reconfigurable, intelligent, and sustainable
wireless environments for 6G smart connectivity,” IEEE Communications Magazine, vol. 59, no. 10, pp. 99–
105, Oct. 2021.
[ADR+22] A. Albanese, F. Devoti, V. Sciancalepore, M. Di Renzo and X. Costa-Pérez, "MARISA: A
Selfconfiguring Metasurfaces Absorption and Reflection Solution Towards 6G," IEEE INFOCOM 2022 - IEEE
Conference on Computer Communications, London, United Kingdom, 2022, pp. 250-259, doi:
10.1109/INFOCOM48880.2022.9796976
[HEXAD3.3] Hexa-X Deliverable D3.3 “Initial analysis of architectural enablers and framework”
https://hexa-x-ii.eu/wp-content/uploads/2024/04/Hexa-X-II_D3.3_v1.0.pdf, April 30, 2024
[OPTI24-D21] OPTI-6G Deliverable D2.1. Cosmas, J., Koffman, I., Lartigue, C., Béchardergue, B., Kapovits, A.,
& Bienvenu, A. (2024). D2.1 OWC Cell Free Network Use Cases and User, Functional and Technical
Requirements (1.0). Zenodo. https://doi.org/10.5281/zenodo.14654936
[OPTI24-D41] OPTI-6G Deliverable D4.1. Koffman, I., Globen, B., Cosmas, J., Mahbas, A., Ali, K., Lartigue, C.,
& Béchadergue, B. (2025). D4.1: OWC Localization Sensing System Design (1.0). Zenodo.
https://doi.org/10.5281/zenodo.14851632
[C+25] E. C. Strinati et al., "Toward Distributed and Intelligent Integrated Sensing and Communications for
6G Networks," in IEEE Wireless Communications, vol. 32, no. 1, pp. 60-67, February 2025, doi:
10.1109/MWC.001.2400056.

Dissemination level: Public | 102

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

7 SECURITY, RESILIENCY, PRIVACY AND

TRUSTWORTHINESS

7.1 SECURITY CONCERNS INTRODUCED BY 6G

The advent of 6G incorporates new paradigms especially focused on a greater capillarization
of services. This new offering encompasses mass sensorisation, ubiquitous connectivity provision
with new types of connectivity, native AI integration, network exposure and programmability
enabling coordination between different functions, and new forms of business focused on
coordination between different providers. As with all other generations of mobile networks, these
new expansions bring multiple ambitions, as well as threats and risks. Vulnerabilities, as yet
unaddressed in its predecessor, are accentuated by the magnitude of data that will be collected
from all parts of the infrastructure. In this respect, security becomes an even more critical issue.
Trustworthiness is one of the main properties on which security efforts are being focused on this
early phase of 6G. Establishing a solid foundation on which to build secure and resilient solutions,
which ensure that properties such as privacy or software genuineness are stable and reliable.
Besides, security solutions must cover complex aspects, and more than ever, and vulnerabilities
can impact the system in such a way, that endanger the whole system safety, for instance by
poisoning data leading into non-intended learning. To this aim, empowering resiliency has
become principal to face security challenges, providing autonomous reaction to threats, even
leveraging AI skins to prevent breaches in advance. In security concerns surroundings, privacy
has been one of the principal topics of 5G, which increases its importance when multistakeholder
and ubiquitous connectivity is added into the scene.

Correct addressing of these concerns starts with the identification of the threat vector and
weaknesses newly incorporated to the landscape. With the whole 6G security picture, addressing
enablers and technologies are studied to establish synergies and build architectural blocks that
supports the main 6G architectural functions by adding security from the design

To this end, we firstly introduce the main threats of the groups identified for the IMT’s six
usage scenarios (i) extension from IMT-2020 (5G) including immersive communication, massive
communication, and hyper reliable & low latency communication; (ii) ubiquitous connectivity;
(ii) AI and Communication; (iv) and Integrated Sensing and Communication:

IMT-2020 extension

The extension of IMT-2020 to meet the needs of 6G has raised a number of security issues.
Rigidly structured and poorly bonded networks find it difficult to adjust to the dynamicity of the

Dissemination level: Public | 103

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

environment, especially in terms of security. The adoption of post-quantum cryptography (PQC),

which is still under study, may lead to signalling overload and new, as yet unknown attack vectors,
leaving the network vulnerable to unforeseen threats. In addition, the increasing reliance on
cloud-native architectures, virtualisation and multi-site orchestration introduce complexity and
vulnerabilities, such as insufficient isolation, miss-configured systems and unsecured APIs - the
perfect testing scenario for an attacker to attempt to disrupt services or gain unauthorised access.

Ubiquitous Connectivity

A key component of 6G networks is ubiquitous connectivity, where devices will be able to

connect from anywhere, inevitably widening the attack surface and increasing the likelihood of
exploitation. While facilitating seamless communication, the constant generation and
transmission of huge volumes of sensitive data raises privacy concerns. Since any device or node
can be an attacker's access point, it is difficult to effectively secure this vast distributed
environment.

Integrated Sensing and Communication

ISAC combines communication and detection capabilities, but doing so carries significant risks.
Sensitive information, such as location and identity, could be exposed due to inadequate data
fusion procedures, with a lack of obfuscation. Moreover, as hackers can exploit unsecured data
flows, the huge volumes of data produced by IoT and ubiquitous networks pose privacy risks. A
security breach at this level can compromise further processes in the system, so there is a need
to maintain the security and accuracy of the merged data.

Artificial Intelligence

Risks associated with AI integration in 6G networks include adversarial AI, where attackers
alter AI models to generate unreliable results and lead decision processes to intentional failures.
AI exploitation, where malicious actors use AI to bypass detection measures, allowing them to
make much more sophisticated attacks. In addition, it is important to note the problem of data
poisoning, which occurs when compromised training datasets compromise network security and
inadequately train AI models or trick analysis engines into believing that unrealistic situations
are occurring.

Network Exposure & Programmability

6G networks are more susceptible to attacks, based on distributed denial of service (DDoS) or
man-in-the-middle (MITM) attacks and data breaches, due to their programmability and
exposure to external environments, as this increases the distributed nature by adding new attack
points accessible to all. As networks become more open and programmable, attackers can use

Dissemination level: Public | 104

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

flaws in software-defined network functions, service interfaces and APIs to intercept private
information, compromising the integrity of communications or drying up processes that overload
network or compute resources.

New Business models

Finally, it is not included in the groups mentioned but it is important to highlight the complex
trust management issues arise between operators and stakeholders as a result of emerging
business models in 6G networks, particularly in shared infrastructure scenarios. Possible risks
associated with the implementation of cloud-native architectures and SLA management include
impersonation, unauthorized access, and in general terms, insufficient authentication,
authorization, and accounting (AAA). Smart pricing techniques could also be exploited to
manipulate invoicing or service access, underlining the necessity for comprehensive security
measures to assure trust and reliability within the ecosystem.

7.2 TRUSTWORTHINESS AS THE PILAR TO BUILD

SAFE, RESILIENT & RELIABLE SECURITY AND
PRIVACY SOLUTIONS
The 6G vision of making 5G an open, multi-operator, user-centric network extends one of the
main security concerns, trust. The incorporation of collaboration with other operators and third
parties to offer advanced services and reuse infrastructures in a transparent way in what is
known as cloud-in-continuum, as well as the softwarisation of IT-based infrastructure makes
trust a central element to ensure throughout the operations cycle.

While in 5G, security architecture follows a centralized network architecture approach and,
trust connections between network parts are created at the protocol level, rather than involving
device and network behaviour. In the envisioned 6G ecosystem, trusted connections are critical
for all parties involved, extending security and privacy to a more inclusive framework, such as
trustworthiness, which should be assured as a native feature.

Figure 7.1: Anchors of trusts & services built upon them extend trust chain

Dissemination level: Public | 105

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Therefore, the most significant paradigm adjustments in the envisioned 6G system are the shift
from a security-only focus to a broader scope of native trustworthiness, clarifying that the term
"trustworthiness" refers to a holistic approach, building safety, security, privacy, resilience and
reliability upon the anchors of trust. Trustworthiness then refers to the solutions coming from
the trusted foundation, platforms and hardware, extending the chain of trust from the basis
towards the user-centric perspective.

Trustworthiness challenge must come with a realistic solution, recognizing all security
measures (i.e. safety, security, privacy, resilience and reliability) come at a cost in terms of
usability, agility, or swiftness. As a result, the envisioned trustworthiness framework should
provide a balance between the various security measures by dealing with a security-by-design
approach, as well as a wide range of themes, such as the trust model and the application of new
cognitive coordination technologies (e.g., Intent-based trustworthiness, based on AI and ML
techniques).

7.2.1 TRUSTWORTHINESS AND LEVEL OF TRUST RELATION

Currently, IMT 2030 promotes trustworthiness as a new attribute for the 6G vision. In fact,
numerous standard groups [ISO16], including 3GPP, ETSI, and IEEE, have been working on
trustworthiness issues. Meanwhile, the world's main communications suppliers have explicitly
underlined the importance of 6G trustworthiness in their 6G projects, proposals, and white
papers. In addition, several scholars have produced technical papers on the definition, generation,
protection, and optimization of trustworthiness. All of these suggests that trustworthiness will
become an essential characteristic in 6G, but still, it remains confusing the use of trustworthiness
and trust terms.

Figure 7.2: Trust and Trustworthiness definitions

Dissemination level: Public | 106

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Despite the discrete meaning and scope of trustworthiness, it is still observed its misuse as
trust. For this reason, it is necessary to start by defining the meaning and relation of
trustworthiness and trust, so it is of highest importance their meaning to be clear and sound.

Trust is an attitude that a tenant has towards a 6G system. In contrast, trustworthiness is a

system property that creates trust to the 6G tenant/user. Thus, a user/tenant trusts (or requires
a specific level of trust from) a 6G system, because the 6G system is trustworthy. In other words,
the trustworthiness of a 6G system contributes to building the trust level of the tenant/user of
the specific system. Thus, the more trustworthy the 6G system is, the higher the trust level of the
tenant/user will be [K+21].

7.2.2 6G PROPERTIES IN TRUSWORTHINESS

User-centric and AI-Assisted Coordination

Considering that each tenant/user has different requirements in terms of trust level from a 6G
system, it is created the need the 6G system to be capable of being adapted to the specific needs
(i.e., trustworthiness level) that corresponds to the level of trust and requirements of each specific
tenant/user. Therefore, the 6G system should not be only trustworthy in a static way, but it should
become user-centric and capable of dynamically adapting the trustworthiness level to the trust
level requirements of each tenant. The user-centric approach in 6G system allows for dedicated
network services to be provided at the granularity of the user by configuring appropriately the
6G system (i.e., the core network functions) [YAZ+23].

In order to achieve user-centric adaptability, intent-based driven trustworthiness is required

for dynamic configuration of6G systems. Both paradigms are necessary to be introduced in the
trustworthy 6G concept, driven by AI/ML-assisted coordination component, which will act as an
intent-handling function that comprehends sophisticated and abstract trust intent semantics
(divided into the five trustworthiness dimensions/taxonomies of Safety, Security, Privacy,
Resilience, and Reliability), and calculates the ideal goal state and organizes activities to
transition the 6G system into this trustworthy state.

Dissemination level: Public | 107

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 7.3: User-centric and AI-assisted coordination of 6G trustworthiness

The AI-assisted coordinator should perform the process of mapping the trust-intent semantics
received by the tenant/user into transition actions of the 6G system via configurations in the
trustworthy dimensions of the 6G system, (i.e., into the safety, security, privacy, resilience, and
reliability domains). Given that this autonomous mapping is a technical application of cognition
(since it is designed to perform the operational tasks of understanding by experiencing and
monitoring), the envisioned AI-assisted coordinator is named Cognitive Coordinator. Besides, the
generation of trust intents shall be allowed to be autonomous. According to the requested service,
a minimum granted trustworthiness should be given with regard to the criticality of data, service,
and networking aspects, such as multistakeholder cooperation. In granting that the system is
protected against malicious users that could try to intentionally degrade the level of trust of a
given service.

Explainability

The use of trustworthiness measures by the system could impact users’ perceived level of
trust, especially if users cannot understand trustworthy actions taken. In this context,
explainability is often viewed as an effective way to build trust among stakeholders. If users have
a better understanding of the process by which the system generates its outputs and the
explanation provided for a particular result aligns with their preconceptions of what constitutes
a proper decision, then the level of trust of the system has been improved. The literature does, in
fact, frequently link explainability to trust [CBS21][Pie11], and many researchers—at least
tacitly—assume that explainability and trust are strongly related [GRR+19] , [RSS16]. This
relationship is known as the Explainability-Trust-Hypothesis, which states that “Explainability is
a suitable means for facilitating trust in a stakeholder.” [LOS+21]

Dissemination level: Public | 108

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 7.4: Explainability as an additional means for improving the Level of Trust 6G

Among the various explainability tools, there is a tendency for modern systems to select
eXplainable AI (XAI) [LOS+21] as part of the to complement the operation of the Cognitive
Coordinator and contribute to the improvement of the tenant’s/user’s Level of Trust over the
Trustworthy 6G System. As evidenced by the "right to explanation" outlined in the General Data
Protection Regulation (GDPR) and the European Commission's (EC) Technical Study on "Ethics
recommendations for trustworthy AI" [KP20], trustworthiness has become crucial for both users
and governmental organizations. They claim that explainability is a crucial element for
trustworthiness. As a result, XAI, or an AI “that creates information or reasoning to make its
working obvious or easy to understand,” is receiving more and more interest from both industry
and academia. In this context, two strategies for achieving explainability can be identified: The
adoption of post-hoc explainability techniques (i.e., the “explaining black-box” strategy) and the
design of inherently interpretable models (i.e., “transparent box design” strategy). These
approaches allow to understand the model behaviour and can be integrated in the model training
or being applied as post-hoc approaches, after ML training.

7.3 CHALLENGES
Adaptative and unattended cybersecurity

Regular 5G orchestration frameworks do not contemplate security as a native element in

service orchestration. In addition, very specific solutions that only work under certain premises
make it difficult to use them in other scenarios. Lacking in adaptability and automation under
heterogeneous infrastructure and conditions. A clear integration of security modules into a ZSM
architecture is mandatory to be able to respond to further 6G security challenges.

Dissemination level: Public | 109

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Following a modular approach, ZSM security framework could be extended to incorporate new
security capabilities and assets, propelling the flexibility to adapt security to novel scenarios and
technologies with an E2E perspective. As well as to have several alternatives to respond and
evaluate threats.

ZSM as a service: AI driven intent management.

5G lacks on coordination between third parties, also third parties’ domains and networks that
would compose a 6G Network Service could potentially lack on self-driving capabilities (e.g. self-
healing). ZSM as a service is a challenging task that aims to provide such autonomous AI based
intent management on demand. This ambitious service involves some other challenges, such as,
the common representation of the system’s exposure functions under an expandable and well-
defined format. Such a system model needs to contain: services running, resources available,
lower-level orchestrators… among others. In this context, to autonomously manage the security
of dynamic topologies is a daunting challenge.

Security SLA (SSLA) negotiation.

Normally, Service Level Agreements (SLAs) have been widely used to represent contract
between customer and providers, but 5G lacks on a common format to model interactions
between different stakeholders and customers. Optimized resource sharing and network service
compositions won’t be possible until communication is automated through a common format.
Defining the standardized model to cover these aspects present some real hurdles due to its
complexity, leading into difficulties for automating the negotiation between the participants as
well as the definition of a baseline to define the expected system and service behaviour.
Furthermore, regarding security, the definition of a common abstraction model used to define the
security aspects of a security service deployment and management forming the Security and SLAs
(SSLAs) is a challenging task.

Resilience and Service continuity

Even though 5G solution proposes dynamic network service deployment, they are normally
realized by statics templates. Dynamic NS composition tailored to SSLA/SLA, possess a significant
challenge specially on how to include multiple domains and grant service duplicity or alternatives
NS to build resilience service delivery.

Application of Quantum Key

Quantum computing is pushing forwards, traditional forms of encryption will become

obsolete, new wave of key generation and encryption resistant to quantum computing styles are
needed, how to integrate the novel PQC in constrained devices. Ensuring seamless and secured

Dissemination level: Public | 110

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

transition to quantum-resistant system introduce unique challenges. Performance overhead of

PQC algorithms, especially lattice-based and hash-based, is one major drawback. Larger key sizes
and therefore increased needs on computational power can put resource-constrained devices like
smartcards and IoT gadgets under stress. This increased demand complicates the deployment of
PQC in environments where high efficiency and low latency are required.

Besides performance-related concerns, migrating to PQC requires planning in ensuring

interoperability between current systems and novel quantum-resistance algorithms. Coexistence
during this phase will happen, hybrid cryptographic schemes are very often adopted, which
embed traditional methods along with the post-quantum ones. This adds complexity for
uninterrupted and secure management operations.

The security of PQC algorithms relies on different mathematical foundations than classical
cryptography. Building confidence in these novel algorithms is an ongoing challenge. Continuous
research and cryptanalysis are necessary to validate these algorithms against evolving quantum
threats [KP20].

The successful adoption of PQC will, in the end, be a function of how industries and
government agencies can effectively incorporate new standards such as FIPS 203, 204, and 205
into their existing cryptographic infrastructures. This would require updating of legacy systems,
ensuring compliance with evolving protocols, and achieving coordination across diverse
stakeholders. The interaction of these challenges underlines the complexity of implementing PQC
and underlines comprehensive strategies for a secure and efficient transition to quantum-safe
technologies.

DLT-based trustworthiness

5G has made significant strides in data speed and service management, but it falls short in
addressing security concerns produced by next-generation connectivity. With the envisioned
heterogeneity of devices and new shapes of service delivery, perimetral-based 5G security do not
cover 6G attack surface. Zero Trust Architecture and its integration with DLT-based solutions are
a challenging task to ensure trustworthiness of the system. The strict access policies provided by
ZTA as well as the verifications needed possess a challenge in energy consumption and scalability.
In this sense, the need for trustworthiness in multi-domain environments becomes a key
challenge, specifically where Multi-Agent Systems take place, enabling dedicated management
but generating the challenging task of enforcing trust between the agents. Integrating such
attestation processes in distributed environments with added difficulty of secured key
distribution. The integration of DLT to solve these challenges also come with scalability problems,

Dissemination level: Public | 111

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

especially to support geographically distributed agents in these multi-domain environments

[DES23].

Continuum of Trust

Since the digitalisation is impacting an ever-increasing part of daily life, connectivity and
supporting network have become a daily necessity and support critical activities and have
become a vital asset. For instance, energy, transport, finance and public administration are
recognised by NIS2 [NIS24] legislative framework as essential entities having to comply with the
constraint of resilience, while heavily relying on telecommunications. Given the importance of the
service they deliver, network infrastructure must be accepted and trusted by their users.

However, as the 6th generation of mobile networks is anticipated arising around 2030 [Eri24],
several fundamental paradigm changes is foreseen hampering the trust in the resources.
Specifically,

The openness of the network, due to the (i) API-ification of the network capabilities to third-
party services and (ii) the increased involvement of diversified solution providers (including
open-source communities). This situation exposes the network to the composition and
interaction with software having a different degree of resilience and robustness.

The involvement of multiple providers, providing and supervising their own network segment
accounting for access network, transport technologies to be dynamically integrated to ensure
constant and optimal connectivity (corresponding to the network of the network paradigm). This
understands the involvement of actors enforcing different security management, which requires
different trust model to cooperate.

The cloudification of the network services and their continuous desegregation throughout the
cloud continuum will increase the dynamism of applications deployment and the providers'
resource usage. This dynamism will make it infeasible to establish a clear boundary for network
service and therefore outwitting any attempts for a perimeter-based security approach.

Cross domain high-scale monitoring & Analysis

Security and privacy enhancement in future 6G networks can be a quite challenging and
demanding task, due to the vast number of potential threats and attacks and their diverse nature
compared to 5G networks (indeed, a larger attack surface is expected in 6G networks). In the
same context, the interconnection of a vast number of devices and the support of heterogeneous
deployments (exploiting the cloud continuum paradigm), which are both key concepts of the 6G
era, might leverage security and privacy concerns, since not all devices will have the capability to
execute advanced security protocols due to their hardware constrained nature. In this context,

Dissemination level: Public | 112

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

artificial intelligence/machine learning (AI/ML) approaches that collect a vast amount of data
from the network to train models that can represent input/output pairs with minimum
performance loss, can leverage security and privacy mitigation via the extraction of abnormal
data patterns and the enforcement of the appropriate actions. Compared to conventional non-ML
detection techniques, ML-based misbehaviour detection provides both a higher detection
accuracy against unknown zero-day attacks, as well as a reduced false detection rate.

The deployment of ML approaches for threat detection and mitigation in the 6G landscape is a
quite challenging procedure, dictated by various key driving factors: i) computational efficiency
of the deployed approaches, ii) identification of multiple and even correlated threats and attacks,
iii) continuous refinement of the ML approaches and knowledge distillation, as well as iv) creation
of multiple network intents per case for network recovery [GNT+24]

Extreme virtualization and softwarization

The trustworthy 6G moves beyond the current NF-centric core network towards a user-centric
evolution of the B5G/6G system over the recently researched edge-cloud-continuum, which is
expected to be the primary option as infrastructure for deploying the softwarised components of
a distributed 6G network. Therefore, for 6G to become the human-centric system of systems
requires significant architectural redesign based on the user-centric (i.e., per-user perspective),
given that the network intrinsically handles the state of each UE or user. A user-centric design is
specifically capable of providing to each user a complete instance of a personalised 6G system
through a user-specific core-network synthesis, supporting for example personal data
management, policy control, session control, and mobility management per-user. These
customized nodes are the so-called User Service Nodes (USNs), while regular centralized core NFs
are defined as Network Service Nodes (NSN).

It is in the transformation of NF-centric to user-centric architectures where the paradigm shift

in establishing, maintaining, and scaling network trustworthiness occurs for 6G. Such
architectural evolution introduces several challenges concerning trustworthiness assurance in a
personalized, user-centric network environment.

The challenges in designing a robust trust model that accounts for the dynamic and
heterogeneous nature of USNs lie in how each USN is responsible for the implementation of user-
specific services, policies, and mobility management, often tailored in real time to the evolving
trustworthiness levels and requirements of the user. How would the system ensure mutual trust
between the user and the network, considering when the network becomes distributed,
personalized, and softwarized?

Dissemination level: Public | 113

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

For example, challenges like personal data sovereignty, secure multi-tenancy, and real-time
computation of trust for individual user sessions have to be solved without sacrificing scalability.
Also, embedding trust functions like safety, resilience, and reliability across the NSN-USN
continuum introduces potential trade-offs between personalization, performance, and security.
It will be exciting and also critical to create a scalable and adaptive trust framework for USNs that
is capable of guaranteeing integrity and authenticity while considering user-controlled data
ownership. How could this framework balance user control, adaptability of the network, and
intrinsic trust in this highly personalized environment?

Security of software, virtualized environments and hardware-based platforms

Building trustworthiness on 6G relies on the use of anchor of trust and applications that extend
them to build a chain of trust towards higher layers. Guaranteeing that software deployed is
trustworthy is a challenge in which the integration between trusted execution environments
(TEEs) and attestation form part of the addressing approach. Still, interoperability, scalability and
lifecycle management are challenging tasks related to the integration of both. As TEEs are
growing attention for 6G networks, finding practical solutions is a key element.

Scalability is one of the main concerns. Attesting thousands of applications and TEE instances
in real-time may result in a system overloading. As well as, properly ensuring isolation in multi-
tenant environments, where different users share the same hardware functions. In this context,
preventing malicious actors from exploiting shared resources and vulnerable separation of
processes and data is of the utmost importance.

Life-cycle management is equally important. Initializing, updating, and decommissioning TEEs

and trusted applications need to be handled prioritizing trustworthiness, protecting sensitive
data from leaks. This becomes intricate specially during workloads migration, where other
security vulnerabilities could emerge. In addition to this challenge, the diversity of hardware
platforms, including Intel SGX, AMD SEV, and Arm TrustZone, requires a unified approach to
ensure interoperability and security across different implementations.

Privacy preserving approaches

The research developments towards the vision of 6G networks represent a substantial

advancement in communication technology, for significant improvements in connectivity, speed,
and innovation. However, this progression also introduces security and privacy challenges. As 6G
integrates an expansive network of devices and services, protecting sensitive information
becomes paramount. Traditional security frameworks are inadequate to address the complex
threats and privacy risks inherent in 6G ecosystems [NLC+21].

Dissemination level: Public | 114

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

At the same time, Privacy is considered a key pillar in EU research and development activities
towards 6G, as privacy enablement is considered a top societal aspect in the EU 6G vision
[BCG+24a]. 6G is envisaged to comprise a decentralized, zero-trust, globally connected
continuum of heterogeneous environments involving several actors across the service chain
(core/edge/RAN infrastructure providers, service providers). In such a pluralistic environment,
privacy is pivotal, not only for the end users but also for all involved stakeholders; and it needs to
be considered as a critical requirement in all technologies of the network stack, including security
mechanisms.

In other words, the challenge for security enablers in future networks is, on the one hand, to
address the significantly widened 6G threat landscape, while, on the other hand, to preserve the
privacy of all actors in the 6G chain. Intrusive security cannot be anymore considered acceptable.

Identity management

The forthcoming 6G networks are expected to be accompanied by extensive collaboration

between stakeholders from different domains. From large infrastructure providers to specialised
microservices. This cooperation will enable transparent services to be offered to a large number
of users of all types, irrespective of the subscriber's home operator. This development brings
many advantages, such as offering advanced, ubiquitous, resource- and price-optimised services,
but it also introduces major security challenges. The heterogeneity of stakeholders and customers
hampers trust management and security arrangements, exposing part of the root of the challenge
in having a compatible AAA system for all participants. Authentication, Authorization and
Accounting (AAA) between stakeholders and users must be fully reliable and trustworthy. To
address the security challenges in 6G, ensuring user privacy at all stages of service provisioning
is of the upmost importance. To this aim, given the vast offer of services and participants, fine-
grained permissions must ensure the subscriber only displays the minimal necessary permission
in each request, possibly through the use of attribute-based access control schemes (ABAC). For
further emphasis, 6G network will deal with sensitive and high-data volume user data, therefore
protecting privacy requires not just minimizing the data exposure but also ensuring unlikability
across the service chain and correlated metadata analysis. In this context, traditional AAA
systems do not embrace the decentralized nature of 6G [BCG+24b]. New wave of mechanisms,
integrating cryptographic methods with DLT and Decentralized Identity (DiD), can offer more
robust and tamper-proof domain-less authentication to the decentralized services and
infrastructure, also providing non-repudiation of actions and transactions occurred. But always
bearing in mind that the use of these solutions entails a notable increase in energy consumption
and scalability challenges.

Automation of Federation

Dissemination level: Public | 115

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Orchestration has been widely applied in recent years, encompassing several layers, trying to
cover every single domain and possible tasks of cellular networks. But still, because of its
increasing complexity, security is a major challenge usually omitted in orchestration studies.
Massive incorporation of devices of all kinds, as well as the integration of new or evolved
technologies, in the 6G picture expand the attack surface to unknown limits. Massive sensor data
collection and processing is a clear example. Interactions between orchestrators of different
levels and cross domains represent another challenge, as each domain may have its own security
protocols, assets, and physical devices. The lack of a common model to format data and coordinate
the exchange of information expose a handicap of current networks. The absence of such a model,
not only complicates coordination but also limits the information sharing, particularly important
in terms of security. Besides, conflict and dependency detection and resolution are a challenging
problem that accentuate the difficulties of the coordination. Described security orchestration
challenges form a subset of challenges inherit from the Zero-touch Service and network
Management (ZSM) paradigm. In this sense, the hierarchical view of ZSM needs to be updated
integrating novel approaches that enable the horizontal learning and action enforcement. The use
of federated agents in ZSM to share knowledge and information on security management tasks
could help decompress this challenge. Orchestrating the interaction between FL agents while
ensuring privacy-preserving Cyber Threat Intelligence sharing introduces even more complexity.
Each agent may use different models and security protocols, making it difficult to align the goals
of privacy, security, and efficiency [Z+24]. Besides, the dynamic nature of 6G deployments makes
the monitoring system a challenging task, where the criticality of data flow in real time for
mitigation and decision-making urgent for a correlation between network topologies and
information gathered. Following the cross-domain federation approach, dedicated agents can
minimize delays in coordination and control. However, this distributed approach introduces
several security challenges, making multi-agent systems (MAS) vulnerable to attacks targeting (i)
agent integrity and (ii) inter-agent communication [Z+24].

7.4 OVERARCHING BLOCKS

The keystone for the security and system management in current and future network is the
Security, Orchestration, Administration, and Response (SOAR) closed loops. SOAR avoid the
definition of complex E2E protocols, but rather specify how system components, grouped in
different categories, should be interconnected to accomplish a specific policy/intent. SOAR
specifies how the following layers are interconnected:

Dissemination level: Public | 116

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 7.5: Overarching Security Blocks

7.4.1 TRUSTWORTHINESS

At the heart of the 6G security architecture lies the Trustworthiness pillar, which serves as the
foundational motivation driving the design, implementation and application of all other pillars.
Trustworthiness, is divided in several domains, including Safety, Security, Reliability, Resilience,
Privacy and Trust. Therefore, Trustworthiness solutions lie on one or more of these domains.
Building trustworthiness is of the upmost importance, as confidence from customers and
stakeholders relies on it. Trustworthiness must be built upon the anchors of trusts, which means
effectiveness of applied domain properties can be proof. Thus, trustworthiness can be measured
maximizing the trust posture over different targets, for instance between administrative
domains, customers or final service consumers. Therefore, applying trustworthiness solutions in
the different fields and layers, facilitate the collaboration for cohesive and unified security

Dissemination level: Public | 117

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

strategies. As mentioned, trustworthiness is present in the rest of the layers as the driving factor
to apply measurements, but in particular, we identify several efforts to apply it through:

• Policy Administration and Enforcement Inspection [iTrust]: Specifies intent-based

security and trust policies while providing explanations for their usage. This pursuit
transparency and confidence in the system, as allow to stakeholders to understand the
implications of the intent in the system.

• Static and Dynamic Trust Assessment [iTrust]: capability to conduct evaluation of

posture of assets depending on their design (e.g. conformity to referential), their current
posture (e.g. resource integrity and behaviour), and support forensic evidence.

• Collaborative Cyber Threat Intelligence Sharing [iTrust]: Establishes a baseline for

trust and security-oriented collaboration among multiple providers. It leverages
standards such as OASIS STIX, TAXI or MISP to enrich information on threat to account
for trust context.

• DLT-Based Trust Infrastructure [Privateer, Desire6G]: By using one of the trusted

anchors, it facilitates secure data exchange with transparency, traceability and
accountability.

• Privacy-Aware Orchestration [Privateer]: which manages network services in

compliance with privacy regulations like GDPR, guided by Level of Trust (LoT)
assessments.

• Proof-of-Transit Mechanism [Privateer]: Ensures data flows through secure,

predefined paths to safeguard privacy against potential attacks.

• Trust Exposure Layer [iTrust]: Limits shared information with external entities,
exposing only necessary trust metrics without revealing sensitive infrastructure details.

• Privacy-Granting AAA Management: Manages the identity cross-domain and for third
party services in a privacy preserving approach through 3 main components:

o Issuer: Central authority for certificate creation using permissioned blockchain

to generate and register subscriber’s DiD and Verifiable Credentials (VCs).

o Subscriber’s Wallet: Empowers users with Self-Sovereign Identity (SSI)

principles, allowing full control over ID management and minimizing information
exposure through pseudo-attributes based on p-ABC schemes.

Dissemination level: Public | 118

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

o Verifiers: Distributed across the service infrastructure to manage access requests

by validating pseudo-attributes without revealing original identities, ensuring
trustworthy service access.

7.4.2 DESIGN LAYER

With a clear integration of human in the loop, and the objective of providing security
modelling, covering from the onboarding specifications for devices and applications in a trusted
way to the policy modelling with different levels of abstraction, enabling the scalability and
interaction between domain’s orchestrators. Principally this pillar are the mechanisms that
define how network components are structured, interconnected, and managed to uphold
trustworthiness. SOAR uses intents that belong to one trustworthiness domain to drive the
interactions and applications, setting a groundwork for a secure and trustworthy network
architecture. Main functional blocks for this block are:

• Intent-based modelling: Defining the models used for intents to define, desired system
status through security objectives and also more specific definition for protection
mechanisms, such as technologies or assets to be used. This block guarantee that security
states are clearly articulated and understandable by domain’s security orchestrators,
propelling scalability and adaptability across diverse domains.

• Trustworthiness Onboarding Specification: Specifies onboarding requirements that

align with trustworthiness domains, ensuring that all assets meet predefined security
standards before being up and running, and accessible. This involves the trusted
integration of devices and applications into the network.

• Human-Controllable UI: Incorporates a human-readable and manageable interface to

manage aspects of the system, such as explainability of security AI processes. As well as
allow overseeing enforced policies, or build new ones. This functional block is the main
interface for enabling human-in-the loop.

• Service Composition: Defines how network services can be chained, and configurations
can be applied based on the security intents. It reduces the complexity of orchestration
and allows for different stakeholders to agree on how their domains will be integrated.

7.4.3 OPERATION LAYER

Receive intents from the design layers or alerts from the analysis assets. Interact with the
AI/ML Layer to drive the enforcement of the high-level intent in the optimal way, considering
several factors such as the network/device conditions and security requirements specified in the
intent. This conforms a Cognitive MANO that through the security orchestrator and the intent-

Dissemination level: Public | 119

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

based security management, reproduces and communicate actions in two ways, to the twinning
deployment to evaluate impact of actions and search for better solutions, and to the real
infrastructural domain, composing and reconfiguring service compositions.

• Security Closed-Loop Automation: Implements intent-based automated set of

recurrent steps interconnecting different logical entities, deployed assets and
infrastructure producing a feedback loop that continuously monitor and adjust security
measures. It ensures that the network remains resilient against evolving threats by
automating responses based on triggered security intents and service compositions.

• Automated Software (SW) Creation and Validation: Leverages automation to develop

and validate security-related software components. Trusted libraries as well as a set of
validation tests described for the domain that the SW belongs to. This automation
accelerates the deployment of security measures while ensuring their reliability and
proper performance.

• Zero Trust Management: Enforces Zero-Trust principles by managing access controls

and continuously verifying assets. Thus, maintaining minimal trust assumption,
adequation isolation methods.

• Federation: This functional block is a key component with the envisioned horizontal
nature for domain composition in the continuum context for 6G networks. The
management of federated agents that enables CTI sharing and cross-domain security
coordination.

• Management and Orchestration (MANO): Orchestrators mesh is envisioned for 6G.

Orchestrators for different levels and tasks are envisioned to work together in protocol-
less manner, for instance, ML orchestrators that manages the life-cycle operation of ML
model (MLFO), orchestrators focused exclusively on applying security offering Security
as a Service or effectively applying federation.

7.4.4 AI/ML LAYER

Represents the reaction plane, where security mitigations and service compositions are first
evaluated in a Digital Twin infrastructure, allowing the impact evaluation and recalculation of
intents enforcement through the virtual infrastructure. Monitoring and Analysis, in which we
highlight the need of a topology inventory agent, tracking the network topology and providing
updated information for threat detection and mitigation planning. Also, the network security
analysis, mainly classifying data traffic from genuine or anomalous, by extracting and analysing

Dissemination level: Public | 120

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

metadata. This component can take into consideration the federation of agents, for data and
analysis correlation.

• Twinning: The network DT acts as a dynamic representation of the mobile network,

constantly learning and evolving alongside the real network environment. ML algorithms,
within this DT framework, can leverage historical data, network topologies, and user
behaviour patterns to model normal network behaviour and promptly identify deviations
that may indicate malicious activities. This integrated approach not only enhances the
precision of threat detection but also empowers security systems to both anticipate and
proactively mitigate potential risks as well as analyse the impact of any proactive action
to be taken.

• Learning: Processes enabling the federated learning between distributed agents and
models. This module is responsible for the distributed ML training. To this end, privacy
preserving solutions are leveraged, such as federated learning. Before the actual training
takes place, preprocessing and feature preparation takes place. Trained ML models are
stored in a local database, where they can be retrieved on demand. All procedures are
orchestrated by the machine learning function orchestrator (MLFO). Adversarial training
is used to protect AI/ML models against attacks, while privacy-preserving mechanisms
are integrated to protect heterogeneous data types

• Model Repository: Place in which to hold models, accessible by stakeholders and

orchestration processes. This enables a place to share trained models, selecting optimal
models given the infrastructure and assets used in the service composition. Allows for on-
demand retrieval and deployment of the latest threat detection models, ensuring they are
readily available when needed.

• Decision: 6G Leverages AI-driven decision-making processes to interpret data from

various sources, decision is driven by one or more of trustworthiness domains, such as
privacy-sensitive decisions or trust-based decisions. Prioritising giving trustworthiness
properties to the system while maintaining consistency and reliability. Decisions retrieve
analytics, topology, assets etc. Which complexity lead into having specialized engines that
manages concrete tasks instead of a general decision-making engine. Decisions are
envisioned to target first DT domain to evaluate impact and look for better solutions.

• Analysis: Analytics based on AI and machine learning models also encompasses detection
of anomalies in user equipment (UE) and the network, processing data in a federated
manner to enhance privacy, and following the federated NWDAF deployment scenario of
the 3GPP.

Dissemination level: Public | 121

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

7.4.5 SENSING

Sensing is another pillar to cover over-architectural security. Sensing in 6G extends from the
centralised infrastructure to the far-edge devices. Where security capabilities must be able to
extract metrics from the latter nodes, which are usually the most vulnerable as they are generally
dynamic and mobile in nature. Sensing is one of the main enablers of the cloud in continuum for
these devices, since it is necessary to know their state before forming a chain of services using
them. Sensing is also of particular importance to detect threats outside the security perimeter of
the operators. Sensing itself do not represent a security mechanism but enable new ways of
studying security for early detection and correlation. On the other hand, emerging sensing
capabilities can be used by attackers to incorporate new layer of information from the
surroundings, for instance doing a mobile network mapping, inferring what kind of devices are
connected to the network or also their location. Beside the new threats incorporated by sensing,
it is the main enabler for the following functions:

• Physical Layer Anomaly Detection: Detection capabilities can be used to detect threats
at the physical layer, such as jamming, spoofing, eavesdropping, semantic communication
inference, denial of service attacks and more. By analysing environmental metrics such as
signal patterns, jamming levels and anomalous behaviour at the physical layer, this
module facilitates real-time identification of attacks that could compromise the security
level of the network by threatening communications integrity or degrading network
performance.

• External Perimeter Surveillance: Detection extends security capabilities beyond the

traditional network perimeter by enabling detection of external threats. This includes the
identification of entities not authorised to broadcast on certain frequencies, malicious
devices or environmental disturbances near critical infrastructure. By continuously
monitoring the external environment, this module improves proactive threat detection
and overall network resilience.

7.5 REFERENCES
[ISO16] ISO Central Secretary, “Systems and software engineering – Systems and software quality
requirements and evaluation (SQuaRE) – Measurement of quality in use,” International Organization for
Standardization, Standard ISO/IEC 25022, 2016. [Online]. Available:
https://www.iso.org/standard/35746.html
[K+21] L. Kastner, et al., "On the Relation of Trust and Explainability: Why to Engineer for Trustworthiness,"
in 2021 IEEE 29th International Requirements Engineering Conference Workshops (REW), Notre Dame,
IN, USA, 2021 pp. 169-175. doi: 10.1109/REW53955.2021.00031
[YAZ+23] X. Yan, X. An, W. Ye, M. Zhao, Y. Xi and J. Wu, "User-Centric Network Architecture Design for 6G
Mobile Communication Systems," 2023 Joint European Conference on Networks and Communications & 6G
Summit (EuCNC/6G Summit), Gothenburg, Sweden, 2023, pp. 305-310, doi:

Dissemination level: Public | 122

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

10.1109/EuCNC/6GSummit58263.2023.10188283.
[CBS21] L. Chazette, W. Brunotte, and T. Speith, “Exploring explainability: A definition, a model, and a
knowledge catalogue,” in IEEE 29th International Requirements Engineering Conference (RE). IEEE, 2021.
[Pie11] W. Pieters, “Explanation and trust: What to tell the user in security and AI?” Ethics and Information
Technology, vol. 13, no. 1, pp. 53–64, 2011.
[GRR+19] R. Guidotti, A. Monreale, S. Ruggieri, F. Turini, F. Giannotti, and D. Pedreschi, “A survey of methods
for explaining black box models,” ACM Computing Surveys, vol. 51, no. 5, pp. 1–42, 2019.
[RSS16] M. T. Ribeiro, S. Singh, and C. Guestrin, “‘Why Should I Trust You?’: Explaining the predictions of
any classifier,” in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery
and Data Mining. New York, NY, USA: Association for Computing Machinery, 2016, pp. 1135–1144.
[LOS+21] M. Langer, D. Oster, T. Speith, H. Hermanns, L. Kastner, E. Schmidt, A. Sesing, and K. Baum, “What
do we want from explainable artificial intelligence (XAI)? – A stakeholder perspective on XAI and a
conceptual model guiding interdisciplinary XAI research,” Articifial Intelligence, vol. 296, 2021.
[KP20] M. Kumar and P. Pattnaik, "Post Quantum Cryptography(PQC) - An overview: (Invited Paper)," 2020
IEEE High Performance Extreme Computing Conference (HPEC), Waltham, MA, USA, 2020, pp. 1-9, doi:
10.1109/HPEC43674.2020.9286147.
[DES23] DESIRE6G, “D3.1: Initial report on the intelligent and secure management, orchestration, and
control platform”, 2023. Online: https://zenodo.org/records/10356033
[NIS24] “NIS 2 Directive.” Accessed: Aug. 06, 2024. [Online]. Available: https://www.nis-2-directive.com/
[Eri24] “6G standardization – an overview of timeline and high-level technology principles”, Ericsson,
March 22, 2024. [Online] Available (Accessed: Sept. 13, 2024):
https://www.ericsson.com/en/blog/2024/3/6g-standardization-timeline-and-technology-principles
[GNT+24] P. Gkonis, N. Nomikos, P. Trakadas, L. Sarakis, G. Xylouris, X. Masip- Bruin, and J. Martrat,
“Leveraging network data analytics function and machine learning for data collection, resource
optimization, security and privacy in 6g networks,” IEEE Access, pp. 1–1, 2024.
[NLC+21] V.-L. Nguyen, P.-C. Lin, B.-C. Cheng, R.-H. Hwang, and Y.-D. Lin, “Security and Privacy for 6G: A
Survey on Prospective Technologies and Challenges,” IEEE Commun. Surv. Tutor., vol. 23, no. 4, pp. 2384–
2428, 2021, doi: 10.1109/COMST.2021.3108618
[BCG+24a] Jose Manuel Bernabé, Eduardo Cánovas, Jesus Garcia-Rodriguez, Alejandro M.Zarca, Antonio
Skarmeta (2024), "Decentralised Identity Management solution for zero-trust multi-domain Computing
Continuum frameworks”, Future Generation Computer Systems,
https://doi.org/10.1016/j.future.2024.08.003
[BCG+24b] Bernabé Murcia, J. M., Cánovas, E., García-Rodríguez, J., M. Zarca, A., & Skarmeta, A. (2025).
Decentralised Identity Management solution for zero-trust multi-domain Computing Continuum
frameworks. Future Generations Computer Systems: FGCS, 162(107479), 107479.
doi:10.1016/j.future.2024.08.003
[Z+24] A. Zahir et al., “Distributed Genuine Intelligence: From Agent Integrity to Secure Inter-Agent
Communications,” European Conf. on Network and Communication (EuCNC), 2024.

Dissemination level: Public | 123

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

8 SUSTAINABILITY

8.1 INTRODUCTION
As we advance toward the design and development of 6G, sustainability has emerged as an
essential pillar of next-generation smart networks and services (SNS) and their connectivity
strategies. The digital transformation of our society and economy must align with European
norms and values, including the European Green Deal principles and the global Sustainable
Development Goals (UN SDGs), while ensuring no one is left behind.

The development of more advanced communication networks faces compound challenges:

increased energy demands, complex resource allocation, impacts on biodiversity, e-waste
management, and the rebound effects produced by ICT (referring to the case where the efficiency
improvements in ICT technologies lead to increased overall resource consumption due to higher
usage, offsetting potential environmental benefits), all while meeting heightened social
expectations surrounding digital inclusivity. The SNS JU initiative has incorporated sustainability
at the forefront of its research and innovation agenda, supporting projects that both minimise
their environmental impact (reduced footprint) and maximise their positive contribution to
sustainability challenges (increased handprint).

Environmental sustainability challenges: a footprint

6G networks promise substantial advancements while presenting significant sustainability

challenges. The network's complex infrastructure, coupled with exponential increases in data
traffic, could dramatically expand its environmental footprint. Key sustainability drivers include
energy consumption, resource use, biodiversity conservation, and electronic waste reduction.
These factors underscore the critical need to minimise the environmental impact of network
infrastructure while supporting responsible digital evolution.

The environmental footprint of 5G and upcoming 6G networks is multifaceted, encompassing

energy consumption, rare resource extraction, and biodiversity impacts. To address these
challenges, SNS JU projects are developing innovative solutions focused on reducing energy and
more general environmental footprint through optimised resource allocation, relay management,
and AI-driven network intelligence.

For example, unlike previous generations, 6G networks will deeply integrate non-terrestrial
(NTN) components, including satellites, aerial networks, and high-density terrestrial
infrastructure. Such full integration proves particularly impactful in bridging the digital divide,
extending connectivity to remote and underserved regions where terrestrial infrastructure

Dissemination level: Public | 124

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

deployment would be both economically and environmentally costly. Furthermore, the NTN
component's adaptable architecture enhances sustainability across the transportation sector,
supporting efficient operations in aeronautical, maritime, railway, and land vehicle systems.

In the field of data transmission, scalable and sustainable optical transport networks are being
designed to handle the massive bandwidth requirements anticipated in 6G networks. Through
optical switching and control protocols, these networks enable efficient data transmission via
multi-granular optical nodes and flexible wavelength allocation. Additionally, applying photonic
solutions to the front-haul, mid-haul and back-haul segments, also known as X-haul, promises to
minimise energy consumption by replacing power-intensive electronic processing with highly
efficient optical solutions.

Societal and economic dimensions: a handprint

While environmental sustainability through enhanced energy efficiency is crucial, the impact
of 6G on societal and economic sustainability is equally important. To ensure sustainability by
design, 6G must address the digital divide, manage the rebound effect, and promote equitable
access to digital resources. SNS JU projects are working to provide affordable, high-quality
connectivity, particularly in underserved and remote areas, thus fostering digital inclusivity. The
potential to impact societal equity extends beyond mere connectivity, encompassing issues of
trust, security, affordability, digital literacy, and infrastructure quality.

However, significant challenges remain. The production and disposal of electronic equipment
for 6G networks may contribute substantially to e-waste, natural resource depletion, and
biodiversity loss. Moreover, as a rebound effect, the infrastructure required to support 6G
systems and related device production is expected to increase Greenhouse Gas (GHG) emissions.
These elements — resource consumption, rebound effect, biodiversity conservation, trust,
security, inclusivity, and affordability — are integral to the larger sustainability conversation
surrounding 6G.

Regulatory framework

Finally, navigating the regulatory landscape of sustainability is critical for the deployment of
6G networks. Compliance with the European Green Deal, UN SDGs, and various national policies
presents both challenges and opportunities. The alignment of 6G strategies with broader
sustainability policies requires careful consideration of trade-offs between accelerated
digitalization, societal acceptance, and environmental responsibility. As 6G networks mature,
achieving compliance will demand a comprehensive approach that addresses technological,
societal, and economic impacts.

Dissemination level: Public | 125

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

This chapter explores the environmental considerations and broader socio-economic

frameworks being addressed by SNS JU projects such as BeGREEN [BeG23], ETHER [ETH23],
FLEX-SCALE [FLE23], HEXA-X-II [HEX23], ORIGAMI [ORI24], PROTEUS-6G [PRO24], 6G-NTN
[NTN23], and 6G4Society [6G4S24], examining how these initiatives contribute to shaping a more
sustainable future for digital infrastructure, examining specific technological innovations for
reducing environmental footprint, analysing solutions for enhancing positive handprint through
digital inclusion and societal benefits, and discussing concrete approaches to regulatory
compliance. Each section provides insights into project achievements, methodologies, and
contributions to building a sustainable 6G ecosystem.

Summary of the chapter

The evolution of B5G and 6G networks is driven by the need for sustainability and efficiency.
Incorporating relay nodes and innovative optical X-haul technologies can significantly reduce
energy consumption and Operational Expenditure (OPEX). These advancements, along with
AI/ML techniques, enhance network performance and ensure sustainable operations. By unifying
terrestrial and non-terrestrial domains, 6G networks aim to create a more energy-efficient and
sustainable wireless infrastructure, addressing the growing demands of mobile data traffic. This
section provides technology options pertaining to sustainability focusing on optimizing different
parts of the network, i.e., the Radio Access Network (RAN), the transport or entirely the End-to-
End (E2E) network.

RAN sustainability advancements: Incorporating relay nodes in future B5G RAN networks
can significantly enhance sustainability. By mitigating signal blocking and increasing coverage in
high-density areas, relay nodes reduce the need for additional base stations, leading to lower
energy consumption and operational costs. This energy efficiency extends to user equipment,
which transmits less power when connected to a relay, thus conserving battery life. Studies have
shown energy savings of up to 90% with the use of relays. Additionally, AI/ML techniques can
optimize relay functionalities, further improving system performance and reducing energy
consumption. Overall, relay nodes contribute to a more sustainable and energy-efficient wireless
network infrastructure.

Transport network advancements: The scalability challenges posed by 6G networks, driven

by high end-user rates and massive small-cell deployments, necessitate innovative optical X-
haul network technologies. These technologies aim to provide ultra-high-speed, energy-
efficient all-optical bypasses, significantly reducing power consumption and enhancing
performance. By integrating multiple optical switching granularities and dynamically adapting
bandwidth, 6G networks can achieve high energy savings, contributing to sustainable and
efficient network operations. In parallel, by enabling dynamic functional splits and passive traffic

Dissemination level: Public | 126

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

distribution elements, these technologies can adapt to varying traffic requirements, leading to
more efficient energy use. This approach not only reduces operational costs but also contributes
to the development of sustainable and energy-efficient network infrastructures.

E2E sustainability advancements: The rapid growth of mobile data traffic is pushing current
networks to their limits, necessitating the development of 6G networks that unify terrestrial
and non-terrestrial domains, i.e., aerial and space layers. This transformation will create a
3D architecture, enhancing link capabilities and resource allocation. To ensure sustainability, 6G
networks need to maximize energy efficiency, reducing Total Cost of Ownership (TCO) and
energy consumption. Strategies for efficient and low-complexity resource allocation should
consider various resource types, i.e., network, compute and storage, as well as related constraints,
enabling real-time decision-making and guaranteeing end-to-end optimality. Overall, integrated
Terrestrial and non-Terrestrial 6G Networks (TN-NTNs) aim to develop sustainable solutions for
user association, traffic routing, and network function placement, ensuring sustainable and
efficient network operations.

8.2 RAN SUSTAINABILITY ADVANCEMENTS

8.2.1 RELAY NODES FOR ENERGY-EFFICIENT RAN

The inclusion of relay nodes in future B5G RAN can be useful for different purposes such as
mitigating signal blocking in millimetre wave deployments or increasing coverage and capacity
in high-density areas, leading to a reduction in the number of BSs to deploy. Besides that, the use
of relays is a cost-efficient option for energy saving as it allows reducing the base station transmit
power consumption and, consequently, it facilitates a reduction of MNOs OPEX. At the same time,
User Equipment (UE) that is connected to a relay also transmits less power, thanks to the better
propagation conditions, thus reducing the UE battery consumption. The power consumption
improvements that can be obtained through the use of relays in a university campus scenario
have been studied in [BeG24], considering indoor relays at different positions and buildings.
Energy savings ranging between 50% and 90% with respect to not using relays were observed,
depending on the power consumption model parameters and the required bit rate, and
improvements in the energy efficiency in a factor around 2.6 were obtained.

The deployment of relay nodes in wireless networks can be done with different types of relays.
The first approach involves the use of fixed relays, where the MNO chooses the position of the
relay as an extension of the currently deployed base stations. The second approach consists of
installing relays within a moving element (Moving Relay), such as a bus or a train, as exemplified
in [NCK20]. The last approach involves equipping UEs with a relay functionality, in a way that
some of the UEs may become relay-UE (RUE) and act as a relay between the BS and neighbour

Dissemination level: Public | 127

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

UEs [PS21]. Standardization for relay support in 5G has been introduced by 3GPP in the so-called
Integrated Access and Backhaul (IAB) technology since Rel-16.

The definition of specific functionalities for the control of the relays is crucial to improve the
system performance and reduce energy consumption. The incorporation of AI/ML techniques in
these relay control functionalities is a key concept to make them more efficient. These relay
functionalities cover different aspects. On the one hand, a coverage hole detection functionality
makes use of a set of collected measurements with the aim of identifying geographical regions
with large traffic demands and poor coverage. On the other hand, a fixed relay placement
functionality is in charge of determining adequate geographical locations to place a new fixed
relay and establishing their initial configuration parameters. Additionally, a candidate RUE
identification functionality aims to identify UEs that can be good candidates to act as relays
between the network and other UEs in their proximity. Finally, a relay activation/deactivation
functionality is in charge of the dynamic activation/deactivation of these relays/RUEs with the
objective of improving the network performance and reducing the energy consumption.

8.2.1.1 ARCHITECTURAL COMPONENTS

Figure 8.1: Relay control components and functionalities

Dissemination level: Public | 128

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 8.1 details the components that support the implementation of the above-mentioned
relay control functionalities. On the one hand, as shown in Figure 8.1, a Relay Control entity placed
at the Service Management and Orchestration (SMO) is in charge of the interaction with the relays
for the collection of the network measurements and the relay reconfigurations through an
extended O1 interface, denoted as O1+. In turn, gNB measurements are sent to the SMO through
the O1 interface. On the other hand, the relay control functionalities are sustained by different
rApps in the non-RT RIC. In particular, the Data Collection (DC) rApp is in charge of the
management of the different processes related to the collection of measurements. Moreover, the
Relay Function Management (RFM) rApp is in charge of the coordination and management of all
the functionalities related to the control of the relays. This RFM rApp decides when and where to
trigger the execution of each functionality. Each AI/ML-based relay control functionality is
sustained by a different AI Engine Assist rApp (AIA rApp), which makes the AI/ML workflow
services of the AI Engine accessible to the non-RT RIC. These AIA rApps cover different aspects
such as data pre-processing, model triggering, performance monitoring of the AI/ML models and
determining the necessity of model updates or model retraining. For each relay control
functionality, the associated AIA rApp triggers the execution of the corresponding AI/ML function
hosted in the AI Engine. These AI/ML functions make use of collected information stored in the
AI Engine datalakes/databases to obtain an output (e.g. the result of a clustering, a prediction or
a recommendation) useful for taking adequate decisions of relay deployment or relay
reconfiguration.

8.2.1.2 SOLUTION

This section describes how the components mentioned above can fulfil the different proposed
relay control functionalities. With certain periodicity, a network monitoring process is run to
identify cells that require the activation of some of the proposed relay control functionalities.
Then, for these bad performing cells, a process of data collection is done. It consists of the
collection of geo-located measurements e.g. based on Radio Resource Control (RRC) UE
measurement reports. Collected measurements are stored in a measurements datalake in the AI
Engine (see Figure 8.1). Then, the RFM rApp, activates the coverage hole detection process that
is managed by an AIA Coverage Hole Detection (AIA CHD) rApp, which triggers the execution of
the process in the AI Engine (see Figure 8.1) making use of the collected measurements stored in
the AI datalake. The result of this process is a characterization of the coverage holes identified in
each cell and is stored in the Coverage Hole Database. After the identification of a coverage hole,
the RFM rApp oversees deciding the most adequate solution to address this problem. One
possible solution is the use of RUEs. For this purpose, the RFM rApp may trigger the process of
Candidate RUE identification, with the objective of identifying UEs that may be good candidates

Dissemination level: Public | 129

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

to become a RUE. In general, UEs with good propagation conditions with its associated BS, a
static/semi-static mobility pattern, a periodical and predictable space-time location and large
session durations may be good candidates to become RUE. The list and characterization of
candidate RUEs for each coverage hole is stored in the Relay database. In case no suitable RUE
has been found to address a specific coverage hole, the RFM rApp may trigger the Fixed Relay
Placement functionality to determine an adequate geographical location to place a new relay. In
case a new fixed relay is deployed, the geographical location and the configuration parameters of
this new relay are added in the Relay database. With the aim of improving the system
performance and reducing the overall energy consumption, both fixed relays and RUEs are
dynamically activated/deactivated depending on the number of users in their surroundings. In
order to do this, the Relay activation/deactivation process makes use of recently collected
measurements and information related to the relays status and uses an AI/ML model to take
smart relay activation/deactivation decisions.

8.3 TRANSPORT NETWORK SUSTAINABILITY

ADVANCEMENTS

8.3.1 OPTICAL TRANSPORT NETWORKS SUPPORTING SUSTAINABLE

CAPACITY SCALING

The target end-user rates and massive small-cell deployments envisioned for 6G, as well as the
rates generated by future immersive AR/VR and holographic services supported by next
generations of FTTH networks, pose scalability challenges to the electronic packet layer in terms
of performance and power consumption. This is especially the case of metro aggregation-core
segments where traffic flows in a totally hierarchical way to and from either WAN transit nodes
or CDN caches. In this case, traffic is not meshed but concentrates into a few core nodes, which
simplifies the task but creates hot spots with huge capacity needs. The challenge is even more
complex if the evolution toward centralization of radio processing functions becomes
mainstream in 6G settings. This situation calls for performing disruptive research on optical X-
haul network technologies of Optical Switching Nodes and Transceiver Interfaces to enable
flexible capacity scaling. Concrete goals set to cope with the 6G vision are: ≥10 Tb/s rate per
optoelectronic interface, ≥1 Pb/s capacity per link (utilizing ultra-wideband (UWB) transmission
and space division multiplexed (SDM) fibre solution) and ≥10 Pb/s throughput per multi-
granular optical node (MG-ON) utilizing new waveband-selective switch (WBSS)). These rates go
beyond the capacities of the conventional C-band employed in DWDM systems. The objective is
providing ultra-high-speed energy-efficient all-optical bypasses to 90% of the traffic destined and
coming from the core, seamlessly integrated with the IP layer by means of a smart control plane.

Dissemination level: Public | 130

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

In this context, the support of multiple optical switching granularities is essential to achieve high
energy savings and dynamically adapt the bandwidth used to the traffic flows served.

8.3.1.1 ARCHITECTURAL COMPONENTS

Figure 8.2: An optical node architecture featuring multiple switching granularities [TPU+24].

Most innovation and changes required in existing standards involves the control plane of the
transport network. Current 5G schemes are in many senses technology-agnostic with respect to
the fixed part of the network. This makes the end-to-end guarantees envisioned for real-time
applications hard to achieve. However, transport networks are going through a revolution
towards disaggregation, openness and programmability that enables unprecedented seamless
integration with IP and e2e 6G services. The control of multi-granular optical nodes combining
wavelength, waveband, and spatial switching is complex given the combinatorial possibilities and

Dissemination level: Public | 131

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

implications in the physical layer of multigranular configurations. The Optical SDN controller,
other than traditional wavelength Routing and Spectrum Assignment (RSA), needs to support
dynamic optical band switching (FLEX-RSA), where an optical-band circuit can be dynamically
configured to accommodate an increasing (or decreasing) number of wavelength channels (over
the same path and assigned to the same band) depending on the traffic needs. Recovery of the
optical band, i.e., an entire band (e.g., C band) or a wide portion of it, is also considered. For
example, in case of a C-band amplification malfunction, the optical band could be recovered over
the S band along the same route, or in case of a soft-failure Quality of Transmission (QoT)
degradation, the spectral window could be changed where QoT is superior.

8.3.1.2 SOLUTION

Achieving the target rates in a cost-effective and energy-efficient way can be achieved through
the utilization of ultra-high bandwidth transceivers employing photonic/plasmonic technologies
and the efficient exploitation of optical spatial and spectral switching (UltraWide-Band Spectral
& Spatial Lanes Multiplexing; UWB/SDM). The target is achieving record energy efficiency (sub-
pJ per switched/transmitted bit) and low cost, enabled by photonic integration and optical
transparency, replacing/bypassing power-hungry and costly electronic processing systems (e.g.,
electronic routers/switches). The Optical Nodes and their Transceiver Interfaces should be
controlled by a Machine Learning-enabled SDN control plane running smart resource allocation
algorithms, which will optimize traffic flow routing across network layers and segments,
improving network QoS (high rates, low latency, high reliability/availability) and low cost/power
consumption, as required by 6G specifications.

Network nodes supporting MG-ON capacities ≥10 Pb/s are envisioned to be capable of
switching at multiple granularities, ranging from individual channels to full fibres through wide
spectral super-channels (flexible bands) withing the Ultra-wide band (UWB) window (1460-
1625 nm, (S, C, and L bands)) and must support Spatial Division Multiplexing (SDM) to generate
flexible reconfigurable add-drop in colour/direction/contention-less (C/D/C) ROADMs.

Control and orchestration should rely on a cloud-native architecture such as ETSI

TeraFlowSDN (TFS) controller. TFS is an open-source cloud-native SDN controller able to scale
the management of a large number of flows. Its modular architecture is based on microservices
making use of containers to isolate the functionalities of the components. Each microservice is
independently deployed and the communication between them occurs through a custom open
interface based on Google Remote Procedure Call (gRPC). TFS needs to be extended with new or
restructured microservices such as optical for multi-granular optical support, end-to-end
orchestrator for packet/optical support, and a ZSM (Zero-touch network and Service
Management)-aligned monitoring-analytics-automation loop. This enables to collect information

Dissemination level: Public | 132

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

related to network services, operations, and devices in real-time, analyse data in real-time to
detect any undesired conditions, and autonomously reconfigure/reoptimize packet and multi-
granular optical networks based on the defined policies for the monitored key performance
indicators.

8.3.2 OPTICAL TRANSPORT NETWORKS SUPPORTING FLEXIBLE

FUNCTIONAL SPLITS

The latest 5G new radio numerology and, extrapolating, 6G radio cell-free massive MIMO
schemes, will boost dramatically the capacity requirements of fronthaul networks. Irrespective
of whether Distributed or Cloud RAN will be a widely deployed technology in 6G, it is clear that
dynamic functional splits are a useful tool to move the radio processing functionality deep into
the MAN in search of reduced energy consumptions (due to the disconnection of elements in the
Distributed RAN) when the network load allows such centralisation. This would not be possible
without a dynamic, flexible, scalable, cost-effective, high-bandwidth, and low-latency optical
transmission and switching technology in the fronthaul network that makes use of passive traffic
distribution elements as we get close to the edge (interfacing the RUs), alternative to high energy
consuming packet switches. Once accomplished, it would provide opportunities to better exploit
the multiplexing gain across multiple RUs by adapting the individual dynamic functional splits to
the varying traffic requirements of all cells fed by a common set of fronthaul/midhaul links.
Achieving such inexpensive technology is extremely complex.

This would not be possible without a suitable agile service management, orchestration and
control system to enable dynamic reconfiguration of the functional splits in the RU/DU/CU and
dynamic reconfiguration of the packet-optical X-haul network to deliver the required transport
capacity demanded by the selected split level.

8.3.2.1 ARCHITECTURAL COMPONENTS

A photonic element that can help to build a fronthaul network to cope with the aforementioned
challenges is some sort of spatially-diverse point-to-multi-point (SDPtMP) optical fronthaul
distribution network as depicted in Figure 8.3. This element is intended to distribute efficiently
the energy (rather than through a splitter) to the edges and it is connected to an advanced ROADM
(Reconfigurable Add-Drop Multiplexer) at a Central Office. Alternative, splitter-based Digital
Subcarrier Multiplexing (DSCM) can perform this same function at a shorter reach, lower speed
but with more flexible bandwidth allocation options.

Dissemination level: Public | 133

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 8.3: Location of spatially-diverse point-to-multipoint devices in the overall X-haul

network.

Innovations require proper standard protocols to realize the flexible functional split concept,
possibly via O-RAN, both from the functionality migration point of view and from the ability to
configure the network to adapt the capacity to the traffic profile according to functional split and
load.

8.3.2.2 SOLUTION

A possible implementation of the aforementioned concept is a spatially-diverse point-to-multi-

point (SDPtMP) optical fronthaul distribution network, with degree-four SDM (Spatial Division
Multiplexing) feed for a first capacity multiplier, and WDM for a second capacity multiplier
[CGG+23]. Introducing an SDM feed to the ODN and establishing capacity allocations in digital
subcarrier allocations per RU and CU, can allow all RUs to operate on a common optical
wavelength without interference. This can be accomplished by optical devices that optically
separates transmitted digital subcarriers using an array of precise optical interleaving filters.
These filters are designed to jointly operate as a circular subcarrier mux/demux. Once the
capacity is exhausted across the SDM group, a second optical wavelength may be introduced and
assigned to particular RUs using a reprogrammable WSS or fixed demultiplexer. The SDPtMP
architecture benefits from SDM/WDM capacity scaling and the assignment of spatial/spectral RU-
CU connections, without the need for splitters/combiners in the light-path.

Next-generation Digital Subcarrier Multiplexing (DSCM) transceivers can be used for dynamic
functional split up to option 7-2. Innovative ultra-high-speed, low-latency, low-cost, and power-
efficient Lite-Coherent (LITE-COH) transceivers (TXR) can be developed as key enablers for the
realization of cell-free MIMO. This can allow for ultra-high fronthaul capacities, as specified in
functional split option 8, enabling 6.4 Tbps per fibre over 8 wavelengths (l). The development of
the LITE-COH TXR will feature 0.8Tb/s per-l, while also contributing to reduced latencies and

Dissemination level: Public | 134

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

significant cost and power-reductions by virtue of all-optical-signal-processing (AOSP)

functionality, which will replace the bandwidth-limited and power-hungry DSP used to process
I/Q quadratures and orthogonal polarizations in conventional coherent TXRs. As a result, the new
TXRs will save an estimated 50% power.

The control plane should also account for the design of data models and protocols of the key
optical network elements to enable fully programmable and monitorable X-haul network
deployment. Autonomous networking architectures and optimization algorithms for efficient
packet-optical network resource, network function, and service management will also be
addressed.

8.4 E2E SUSTAINABILITY ADVANCEMENTS

8.4.1 REAL-TIME SUSTAINABLE RESOURCE ALLOCATION IN

INTEGRATED TN-NTNS

The ever-increasing traffic growth of mobile data is about to push the capabilities of current
networks to their limits in the upcoming years. Considering that, 6th generation (6G) networks
are about to unify the terrestrial and satellite domain so as to offer a wider selection of access
points. Thereby, the transformation of current Radio Access Network (RAN) is anticipated,
leading to a 3D architecture comprising “a network of networks”, with the expanded network
offering even more front/mid/back-haul (X-haul) link capabilities through, e.g., Inter-Satellite
Links (ISLs).

In this context, resource allocation becomes even more challenging, due to the additional space
and aerial base stations (BSs) and X-haul links, requiring new channel modelling to support them.
In parallel, the User Equipment (UE) Service Function Chains (SFCs) and their associated x-
Network Functions (xNFs), which can be network functions of any type, e.g., physical (PNF),
virtual (VNF) or cloud-native (CNF), have to be deployed ensuring that the capacity and
computational constraints of each node are met and that the xNFs are executed in the same order
as in the SFC without violating any link constraints [LGLY21]. xNFs can be placed either to BSs (in
both terrestrial, aerial and space domain) collocated with Multi-Access Edge Computing (MEC)
capabilities for lower latency, or to farther cloud computing nodes with greater capabilities but
with higher latency or to fog computing nodes in between offering a trade-off between computing
capacity and latency. For wireless links, the use of the mmWave band is favoured, while for the
ISL links higher frequencies, i.e., optical connections from 20 to 375 THz [ITU02], will be
preferred exploiting the lack of atmosphere in space.

Dissemination level: Public | 135

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Due to the network densification dictated by the unprecedented data traffic growth, 6G
networks are forced to maximize their energy efficiency, mainly to: a) reduce the associated
Operational Expenditure (OPEX) of the network and b) decrease the energy consumption, thus
leading to sustainable networks. As a result, due to these imminent additions and modifications
to future mobile networks, strategies for online resource allocation should be designed that: i)
consider different resource types, such as computational, communication and storage, as well as
6G technologies, e.g., THz bands, and their constraints, ii) enable real-time decision-making by
optimizing the algorithmic computational complexity, iii) guarantee end-to-end (E2E) optimality
by taking into account the E2E latency and data rate needs and iv) maximize the network energy
efficiency. In a nutshell, integrated Terrestrial and non-Terrestrial 6G Networks (TN-NTNs) call
for the development of energy-friendly solutions for online user association, traffic routing and
xNF placement, while guaranteeing the QoS of the UE and the SFC chaining.

8.4.1.1 ARCHITECTURAL COMPONENTS

For real-time E2E network optimization, 6G networks set forth a vision for end-to-end
Network Intelligence (NI) enabling zero-touch management, which requires the coordination of
many NI algorithms running across schedulers, controllers, and orchestrators, ensuring their
conflict-free and synergic operation. Yet, current frameworks by main Standards Developing
Organizations (SDOs) are far from supporting a native NI integration (e.g., 3GPP Rel. 19 will not
include centralised AI/ML control, which will thus not be considered by the SDO until 2026
[Chu24]). The proposed approach builds instead on proposals by 5G PPP/6G IA [BGG+23], which
are inspired by the results of [DAE25]. A new NI Stratum is introduced for the control of AI/ML
models deployed across the network. The proposed AI Layer (Figure 8.4) is aligned with the
internal organization, AI/ML model representation and operation of the NI Stratum. The AI layer
(internal or AI as a Service, AIaaS) supports the E2E cross-domain Management and Network
Orchestration (MANO) framework to coordinate on multiple levels, which are shown in Figure
8.4, and include:

• Infrastructure Layer – It includes the TN and NTN assets (Core/Central Cloud,

Transport, Edge, Extreme Edge infrastructure), External Infrastructure, Virtualised
Infrastructure, i.e., NFVI and non-virtualised resources.

• Network Layer – Network Functions (NFs) (e.g., 3GPP Core Network (CN), RAN,
transport) and third-party functions

• Service Layer – Network Slice Instances (NSIs) composed of NFs residing in the Network
Layer, slice management and exposure mechanisms.

• Application Layer – applications using functionalities offered by Network Slices (NSs).

Dissemination level: Public | 136

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

• Business Layer – business actors: Mobile Network Operators (MNOs), verticals, etc.

Figure 8.4: Proposed architecture for E2E real-time network optimization.

8.4.1.2 SOLUTION

Both an optimal solution that solves the joint real-time user association, traffic routing and
xNF placement as well as a heuristic algorithm for low complexity are proposed. The optimal
solution is based on the system being modelled as a directed graph, with the set of non-UE nodes
and the set of links among them, solving the joint problem subject to flow conservation, power,
capacity and causality constraints. The heuristic, which is named as Online Power-efficient
Terrestrial Non-terrestrial Resource Allocation Heuristic Algorithm (PETA), studies the joint
problem, aiming to maximize user acceptance ratio as well as energy efficiency. As illustrated in
Figure 8.5, PETA is split into two main steps: a) first the algorithm decides upon the user
association and traffic routing path and then, b) it places the xNFs of the SFC required by each UE,
in the exact order specified by the SFC.

In the first step, every time a new UE SR arrives, PETA constructs a weighted graph and
examines all available paths from the source to the destination based on their power
consumption. In each path, all feasible wireless and fibre X-Haul transport links are included, as
well as the AN link between the serving BS and the UE. The shortest-weighted path, i.e., with the
minimum power consumption, is then selected to satisfy the UE demands, as long as the capacity
and delay constraints are not violated. In case of a violation, PETA selects the next available
shortest path, with no constraint violations and proceeds or, otherwise, if there is no other path
to select, PETA blocks the UE and checks for new SR arrivals.

Dissemination level: Public | 137

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Once a path has been selected, PETA moves to the next step, i.e., the xNF placement. In order
to place each xNF of the requested SFC in the available nodes specified by the selected path, the
nodes are being sorted by a parameter denoted by Ω, which consists of the node’s closeness
centrality, the maximum computational capacity and the load of CPU. As for the CPU load, four
values are allowed: a) 1 (high priority) when the node has enough computational capacity and
can host the xNF without the need for a new xNF instance initiation, b) 0.5 (low priority) when
the node has enough computational capacity but needs a new instance initiation to host the xNF,
c) 0.1 ( very low priority) when the node has enough computational capacity but needs a new
instance initiation to host the xNF and the node was previously inactive, and d) 0 (no priority)
when the node cannot host the xNF. When sorting is finished, PETA selects the highest ranked
node and places the xNF, provided that the computational constraints are satisfied. If such a node
cannot be found, PETA returns to the first step and selects the next shortest path, while repeating
the process in the second step until all xNFs are placed or there are no other available paths, in
which case it blocks the UE. If all xNFs from the SFC are placed, PETA updates the network state
and waits for new SR arrivals, repeating the same steps.

Figure 8.5: PETA’s flowchart

8.5 CONCLUSIONS
This chapter has highlighted the pivotal role of sustainability in shaping the development and
deployment of 6G networks. A high focus has been given on the European projects’ advancements
which mainly focus on how to make the 6G networks sustainable, resource-conscious, and
environmentally responsible by addressing what is called Sustainable 6G. These advancements

Dissemination level: Public | 138

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

may focus on specific parts of the network such as the RAN or the transport or aim to optimize it
E2E in terms of energy-efficiency.

Although energy-efficiency by design has been a key sustainability target among European
projects, its additional role as an enabler, other than a goal, is expected to gain ground during the
next years, towards what is called 6G for Sustainability—leveraging 6G technology to drive
sustainable growth across industries. As we look ahead, challenges like optimizing energy use in
constrained environments and enhancing network efficiency will demand innovative solutions,
with technologies like AI expected to play a significant role. Ultimately, this chapter reinforces the
notion that sustainability will be a foundational pillar of 6G networks, just as 6G will be a vital
enabler of sustainability across the digital ecosystem.

8.6 REFERENCES
[6G4S24] 6G4Society project “Towards a sustainable and accepted 6G for Society”, 2024. Available:
https://6g4society.eu/
[BeG23] BeGreen project “Beyond 5G Artificial Intelligence Assisted Energy Efficient Open Radio Access
Network”, 2023. Available: https://www.sns-begreen.com/
[BeG24] BeGREEN, D2.2, “Evolved Architecture and Power Enhancement Mechanisms”, July 2024, (Online)
Available: https://www.sns-BeGREEN.com/deliverables
[BGG+23] M. K. Bahare et al., “The 6G Architecture Landscape - European perspective.” Feb 2023.
https://doi.org/10.5281/zenodo.7313232
[CGG+23] C. Christofidis, G. Gorgias, H. Georgopoulos, K. Moschopoulos, D. M. Marom and I. Tomkos,
"Spatially-Diverse Point-to-MultiPoint Optical Distribution Network for Enhanced 6G Fronthaul," 2023
International Conference on Photonics in Switching and Computing (PSC), Mantova, Italy, 2023, pp. 1-3,
doi: 10.1109/PSC57974.2023.10297167.
[Chu24] J. Chuang, “AI In Telcom & RAN - Standard Evolution.” Mar 2024.
https://www.linkedin.com/pulse/ai-telcom-ran-standard-evolution-jessica-chuang-0pi4c/
[DAE25] H2020 ICT-52 DAEMON, “Network intelligence for aDAptive and sElf-Learning MObile
Networks”, https://h2020daemon.eu/
[ETH23] ETHER project “ETHER – sElf-evolving terrestrial/non-Terrestrial Hybrid nEtwoRks”, 2023.
Available: https://ether-project.eu/
[FLE23] FLEX-SCALE project “Flexibly Scalable Energy Efficient Networking”, 2023. Available: https://6g-
flexscale.eu/en
[HEX23] HEXA-X-II project “A holistic flagship towards the 6G network platform and system, to inspire
digital transformation, for the world to act together in meeting needs in society and ecosystems with novel
6G services”, 2023. Available: https://hexa-x-ii.eu/
[ITU02] ITU-R S.1590, “Technical and operational characteristics of satellites operating in the range 20-
375 THz,” ITU-R, Tech. Rep., 2002.
[LGLY21] L. Liu et al., “Joint Dynamical VNF Placement and SFC Routing in NFV-Enabled SDNs,” IEEE
Transactions on Network and Service Management, vol. 18, no. 4, pp. 4263–4276, Dec. 2021.
[NCK20] G. Noh, H. Chung and I. Kim, "Mobile Relay Technology for 5G," in IEEE Wireless
Communications, vol. 27, no. 3, pp. 6-7, June 2020, doi: 10.1109/MWC.2020.9116079
[NTN23] 6G-NTN project “6G Non Terrestrial Networks”, 2023. Available: https://6g-ntn.eu/
[ORI24] ORIGAMI project “Optimized Resource Integration and Global Architecture for Mobile

Dissemination level: Public | 139

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Infrastructure for 6G”, 2024. Available: https://sns-origami.eu/

[PRO24] PROTEUS-6G project “Programmable Reconfigurable Optical Transport for Efficiently offering
Unconstrained Services in 6G”, 2024. Available: https://proteus-6g.eu/
[PS21] J. Pérez-Romero and O. Sallent, "Leveraging User Equipment for Radio Access Network
Augmentation," 2021 IEEE Conference on Standards for Communications and Networking (CSCN),
Thessaloniki, Greece, 2021, pp. 83-87, doi: 10.1109/CSCN53733.2021.9686119
[TPU+24] Tomkos, I., Papapavlou, C., Uzunidis, D., Moschopoulos, K., Muñoz, R., Marom, D.M., &
Nazarathy, M. (2024). Towards Multi-Pbps Backbone Optical Networks in Support of Future 6G Networks.
CLEO 2024.

Dissemination level: Public | 140

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

9 NETWORK EXPOSURE CAPABILITIES

Network openness has emerged as a significant integral part of 6G networks. A major enabler
for this potential has been the network exposure capabilities, though the emerging ecosystem of
network Application Programming Interfaces (APIs). Taking advantage of exposure APIs, various
advancements enabled, such as deterministic networking, programmability at data and control
plane, as well as digital representation of the network infrastructure. Overall, a technological and
business osmosis is being conducted, leading to new architectural approaches.

9.1 NETWORK EXPOSURE CAPABILITIES

9.1.1 ENABLING DETERMINISTIC NETWORKING BY EFFICIENTLY

BRIDGING MULTIPLE NETWORK DOMAINS

Resource management and control exposure has been a challenging task for multi-domain
environments. Each domain often has its own control mechanisms and protocols, which makes
the coordination between domains more complex. As network infrastructures evolve toward 6G,
there is an increasing need for a framework that allows dynamic, real-time exposure and
management of deterministic network capabilities across diverse domains. Such frameworks are
indeed the basis to implement pervasive automation mechanisms characterizing the 6G systems.
Focusing on deterministic networks, it is crucial for the end-to-end service automation of flow
management to have at any moment an accurate picture of the status of the deterministic service
parameters e.g., latency/RTT, jitter, data rate etc. in any domain. This challenging objective
requires that the collection and exposure of data must happen in real-time and in a synchronized
manner. Specific elaborations are furthermore required to calculate the impact of local domain
parameters on an E2E Deterministic service provisioned across multiple network technologies.

The proposed architecture builds on IETF DetNet to deliver end-to-end deterministic services
across multi-domain environments, integrating diverse technologies such as IEEE 802.1 TSN, 5G
TSN, and IP-based networks. By unifying control and data planes, the architecture ensures low
latency, minimal jitter, and high reliability across different segments.

• AI-driven Multi-Stakeholder Inter-Domain Control-Plane (AICP). The AICP provides

the intelligence needed to manage deterministic paths across different network domains.
Using AI/ML algorithms, it dynamically allocates resources, predicts network load, and
proactively manages service-level agreements (SLAs). It ensures cross-domain
coordination, abstracting technology-specific complexities and providing a unified
control interface that seamlessly orchestrates deterministic services across diverse

Dissemination level: Public | 141

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

technologies like TSN and 5G. The AICP includes a suite of Management Services (MS) that
handle critical tasks such as Time Synchronization, Path Computation, and Service
Automation. These services ensure that network devices across domains are
synchronized and that deterministic paths are optimally configured. Time-sensitive
applications benefit from precise timing, while path computation ensures deterministic
flows are prioritized across domains. The monitoring task is addressed by this part of the
architecture, where dedicated MSs [1] collect the parameters from the Data Plane (see
point 2) exposing them in real-time so that enabling analytics and decision process for
service assurance.

• Multi-Domain Data Plane (MDP). The MDP ensures that deterministic traffic can
traverse heterogeneous networks while maintaining strict quality-of-service (QoS). It
leverages DetNet Layer-3 capabilities to create deterministic paths using Packet
Replication, Elimination, and Ordering Functions (PREOF), which prevent packet loss and
ensure consistent data delivery across domains. Furthermore, a 3GPP’s PDU Set approach
has been adopted as Data Unit Groups (DUGs) into an IPv6 header solution, enabling
DetNet L3 capabilities across a set of IP packets instead of individual packets within a
service flow only [2]. The MDP handles flow-based routing to guarantee minimal latency,
coordinating closely with the AICP to maintain deterministic paths across different
domains.

• Interoperability Between Domains. A major challenge is ensuring that deterministic

services can be maintained across different network domains. The architecture integrates
gateways (DetNet Extended routers) at domain borders to translate QoS requirements and
service parameters, ensuring that traffic retains its deterministic properties when
transitioning between technologies like TSN and 5G, as well as doubling as NW-TT and
DS-TT Translators between domains. PREOF and DUG mechanisms ensure high reliability
by replicating packets along multiple paths and eliminating duplicates to avoid packet
loss and out-of-order delivery.

• Scalability and Extensibility. The architecture is designed to be modular and extensible,

with model-driven APIs that allow easy integration of new technologies (e.g., future 6G
innovations). This flexibility ensures that deterministic services can scale to meet the
demands of complex, large-scale multi-domain environments without sacrificing
performance.

Dissemination level: Public | 142

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 9.1: System Blueprint of a Multi-Domain Data Plane for Deterministic Networking

9.1.2 API ECOSYSTEM FOR EXPOSURE AND INTERCONNECTION

SERVICES

A wide set of RESTful APIs has emerged to support northbound interface, allowing external
(third party) systems to easily integrate and automate procedures related to configuration,
performance, and fault management. Already a NaaS approach has been described and is being
developed by GSMA, CAMARA, TM Forum, and other fora, identifying three main API types.

• The Service APIs provide a purpose-specific capability to third parties, including

management APIs, allowing the application developer to run certain management
functions from within the application. The CAMARA project has been the major
contributor for the definition, development, and validation of the Service APIs.

• The OAM APIs offer programmable access to Operation, Administration and Management
(OAM) capabilities to facilitate the integration of the Open Gateway NaaS Platform with
portals, marketplaces and other aggregation platforms.

• The Technology-specific APIs refer to operator internal APIs offering programmable

access to telco infrastructure and network, service and IT capabilities. These APIs are
typically defined in standardization bodies (e.g., 3GPP, IETF, ETSI, TM Forum) and cloud
communities (CNCF) and are typically tied to the underlying technology.

SNS JU SoftNet WG, has released a relevant white paper [SoftNet24] where the related API
ecosystem and the emerging capabilities are presented. From the architectural perspective, the
efforts are led by the Operator Platform group (GSMA) where the target is a framework that can
unify the external integration and exposure, allowing operators to offer their services or
collaborate with hyperscalers and other service providers.

Dissemination level: Public | 143

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

9.1.3 EXPOSURE SERVICES TO ENABLE ADVANCED EXTENDED

REALITY APPLICATIONS

The eXtended Reality (XR) ecosystem is still facing network performance, interoperability,
sustainability, and cost barriers when targeting ubiquitous networked services over
heterogeneous environments [Mon24].

Novel modular and standards-compliant architectural innovations can be seamlessly

integrated to B5G – towards 6G - networks for an enhanced and more flexible support for XR
services, by exploiting network exposure, edge computing / federation, and Network as a Service
(NaaS) principle, and by additionally abstracting service developers from requiring an in-depth
knowledge of underlying technologies and systems, and of associated low-level and domain-
specific APIs. Such API services can support Network-assisted Rate Control as well as Edge
Selection and Lifecycle Management.

Network-assisted Rate Control. Current Over-the-Top (OTT) rate adaptation mechanisms in XR

services can lead to unfairness and stability limitations [Lop24]. A new Network-assisted Rate
Control API is envisioned so that an XR Control Plane (CP) Application Function (AF), like an XR
Orchestrator [Fer23, Mon24], can subscribe to network and metrics exposure functions
informing about service-related Quality of Service (QoS) drops or network-level congestion
situations. Upon detection / estimation, two main mitigation actions can be triggered: 1) The XR
CP AF can enforce rate adaptations by client-side or in-cloud XR User plane (UP) Afs, based on
specific recommendations by a network element with an holistic view of the used resources; 2)
The XR CP AF can request Quality of Service (QOS) to the network, e.g., employing Quality on
Demand (QoD) APIs [Mon24], so that the underlying resources (network slices, compute nodes,
etc.) are re-configured accordingly.

Edge Selection and Lifecycle Management API. XR services can exploit Edge Computing
paradigms to offload processing functions from the clients (thus reducing their computational
resources and favouring interoperability and sustainability) and to bring communication
modules from far Cloud to close-by Edge servers. Relevant examples include the instantiation of:
(i) Multipoint Control Units (MCU) [Fer23] so that smart mixing, transcoding and/or forwarding
media functions allow reducing the computational and bandwidth requirements on the client
side; (ii) Remote Renderers [Yer24], so that efficient support for untethered and lightweight (e.g.,
smartphones) XR devices can be provided. In this context, Edge-Cloud APIs [EdgeCloud] allow to
discover the available Edge resources, facilitating the selection of the desired ones based on
specific criteria (e.g., delay, cost, etc.) and managing the lifecycle of the virtualized AFs (e.g., MCUs,
Remote Renderers) to be instantiated in those Edge servers. In addition, the XR CP AF can

Dissemination level: Public | 144

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

subscribe to network exposure functions (e.g., NEF), like User Equipment (UE) mobility detection,
so that Edge migration and Traffic Influence actions can be triggered to further improve
performance, e.g. selection of optimal routing path for lower latency.

9.1.4 EXPOSURE SERVICES TO ENABLE CONNECTED AND

AUTOMATED MOBILITY (CAM)

6G aims to expand the set of supported verticals and provide enhanced capabilities beyond
connectivity. Already, 5G System (5GS) has been built as a modular architecture to support any
vertical running on top in a vertical-agnostic manner. In this context, Connected Automated
Mobility (CAM) vertical services, have emerged as a broad range of services in and around
vehicles, including both safety-related and other services enabled or supported by the 5GS.
However, it is realized that certain verticals (like CAM) have specific and strict requirements.
Thus, although significant progress has been made in supporting verticals, the corresponding
necessary configuration of the network and end-devices is a time-consuming manual process that
requires tight coordination at technical and business levels across the verticals, the vendors, the
network operator, and even the end-users. This hinders not only the greater adoption of 5GS but
also the uptake of novel CAM Use Cases (UCs) and the modernization of existing ones that require
a tighter integration with the underlying network.

Thus, a main objective towards 6G is to open up the reference 5G adv. architecture, and also to
transform it into a vertical-oriented with the necessary interfaces tailored to the CAM UCs that i)
expose network capabilities to verticals, ii) provide vertical-information to the network; iii)
enable verticals to dynamically request and modify certain network aspects in an open,
transparent and easy to use, semi-automated way. This requires dedicated APIs that can act as an
intermediate abstraction layer that translates the complicated 5GS interfaces and services into
easy to consume services accessible by the vertical domain. The experimentation framework and
the main innovations developed in the project are: Multi-access Edge Computing (MEC) with
service continuity support, zero-touch management, multi-connectivity and predictive Quality of
Service (pQoS).

Dissemination level: Public | 145

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 9.2 Key architectural components for enabling Network-aware CAM applications

9.2 PROGRAMMABILITY ENABLING FEATURES

9.2.1 INFRASTRUCTURE MANAGEMENT LAYER

In 6G network services, packet traffic needs to be forwarded through different network

functions (NFs). Each NF has a control plane and a data plane. A new architectural component
called infrastructure management layer (IML) was proposed in [1] to separate concerns of the
packet processing business logic and the infrastructure layer. IML basically acts as a combination
of a Virtualized Infrastructure Manager (VIM) and a hardware abstraction (HAL) layer. IML is
responsible for managing a pool of resources. IML focuses on the deployment and run-time
management of data plane components. An NF data plane component implements the packet
processing logic and can be executed on various targets including smartNICs, ASICs, FPGAs, IPUs
and DPUs, in addition to traditional CPU resources. IML is responsible for selecting the
appropriate target(s) and number of instances to execute the NF data plane and configure the
virtual links between them at deployment time. Virtual links are created by infrastructure
network functions implementing traffic forwarding and routing between NFs. To enable run-time
optimization and hide the underlying optimization from the NF control plane, IML introduces a

Dissemination level: Public | 146

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

control plane proxy using a common northbound API (e.g., P4Runtime [2]) that provides a single-
instance view of the data plane component to the NF control plane. The proxy hides the
underlying data plane optimization like load balancing between multiple data plane instances of
the same NF data plane or offloading heavy hitter users to hardware data planes. To enable the
better utilization of data plane hardware resources, IML has a subcomponent called P4-MTAGG
[3, 4] that is a compiler-based virtualization tool for P4 [5] programmable hardware targets. It
enables the deployment and execution of multiple P4 programs on the same P4 hardware in an
isolated way. The control plane access to the different data plane programs is also isolated by the
IML’s control plane proxy component.

9.2.2 PROGRAMMABLE TRANSCEIVERS IN OPTICAL TRANSPORT

NETWORKS

The sustainable scaling of the capacity, to support 5G+/6G, requires combining Wavelength
Division Multiplexing (WDM) with Space Division Multiplexing (SDM) to exploit the spectral and
the spatial dimension of the fibre (i.e., frequencies, cores, and modes) using multicore fibres
(MCF), multimode fibres (MMF), or combining cores and modes in few-mode multicore fibres
(FM-MCFs), or bundles of SSMFs. A key challenge is to design and produce a transport network
infrastructure able to support beyond 5G and new emerging services, relying on the joint usage
of Multi-Band and SDM, spanning the access, aggregation, and metro/long-haul segments,
supporting the requirements for X-haul, further integrating the packet/optical and computing
layers, and targeting efficient networks in terms of capacity and energy efficiency. In this view, a
converged packet-optical transport is need based on resources, so that drastically reducing the
presence of aggregation routers and O/E/O conversions, capable of removing boundaries
between different network domains and between networks and computing resources.

9.3 NETWORK REPRESENTATION AND FUNCTIONAL

STRUCTURE

9.3.1 NETWORK DIGITAL TWIN

The digital twinning concept brings real time monitoring and prediction capabilities down to
the network infrastructure. Towards 6G, the integration of the so-called Network Digital Twin
(NDT), is expected to operate across three distinct layers: the physical network, the digital
network, and a federated simulation framework.

Dissemination level: Public | 147

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

The physical layer remains consistent with existing network elements, such as User
Equipment (UE), RAN, and core network, while the digital layer introduces a network twin that
allows for dynamic simulation and control.

The digital layer is built upon the ITU-T Y.3090 recommendation [ITU-T-Y3090], which
outlines two core model types, basic and functional models:

• A basic model of a network element is the collection of data describing its properties,
configurations, and operational status, along with any associated algorithms or protocols
used to emulate its dynamics and evolution with time. A basic model of a network is the
aggregation of basic models of network elements, including their physical and logical
relationships and the interactions that occur between them.

• A functional model of a network builds upon basic models, applying advanced processing
techniques, often through AI/ML algorithms, under varying operational scenarios. These
models are designed for specific objectives such as performance optimization, anomaly
detection, or predictive maintenance.

The third layer, the federated simulation framework, enables the coupling of multiple domain-
specific simulators, forming a unified system that allows for large-scale scenario testing. This
framework supports both online and offline NDTs, enabling networks to perform "what-if"
analyses and refine AI-based functions before deploying them in real-world environments. This
is critical for the orchestration of AI-driven services, providing a feedback loop for real-time
performance optimization.

9.3.2 NETWORK ABSTRACTION TO SUPPORT TRIALS

When it comes to network exposure for experimentation purposes, network and compute
infrastructure should facilitate medium- to long-term experiments without frequent manual
reconfiguration and giving experimenters access to internal network configuration. In this
context the Trial Network concept has been introduced [Tso24].

A Trial Network (TN) represents an end-to-end network with physical and virtual
components dedicated to experimentation purposes. The TNs are fully configurable, manageable,
controllable and automatically deployable networks combining virtual, physical, and emulated
resources.

The TN software components are described in a common repository, called in [Tso24] as 6G

Library, which eases an experimenter to perform a modular and automatic deployment of a Trial
Network by selecting on demand the required elements from the library.

Dissemination level: Public | 148

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

The 6G library's objects are curated and designed to serve as the foundational building blocks
for building the Trial Networks. For the library implementation Github serves as a sophisticated
version control system essential for monitoring changes within computer files, primarily
employed in managing source code during software development. Each element within the 6G
library embodies the Everything as a code - EaC philosophy, designed as self-contained unit
equipped with the necessary automations and scripts for deployment within a network and
compute infrastructure.

The architectural blueprint of a component (6G library element) has been meticulously crafted
with a focus on simplicity, clarity, versatility, scalability, and adaptability. In this view, every
element in the library is becoming Trial Network-ready by complying with a specific predefined
toolset (namely Terraform, Ansible, Jenkins, and Ansible) and is hosted under a common folder
structure.

At functional level, in [Tso24] a common API framework is defined for all the components of
the 6G-library that need to interact at application level with third party software / experimenter.

The brain for the realization of the Trial Network Concept is the Trial Network Life-Cycle
Manager (TNLCM). The TNLCM is, within the [Tso24] Architecture, the entity that ensures that
every Trial Network in each platform is accessible and in working order, as well as orchestrates
the necessary actions required for changing the state of a TN when necessary.

Dissemination level: Public | 149

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 9.3 Abstract representation of the components that realise the Trial Networks concept

9.3.3 USER CENTRIC NETWORK STRUCTURING

Network services will be recentred on users, following a user-centric approach, distinguishing

characteristics of the 6G architecture, enabling user-definition, user-configuration, and user-
control. The user-centric architecture in 6G will alter how users, network services, and apps now
communicate, which will have an influence on the ownership of personal digital assets, network
access, and mobility management. The softwarized telecom service enables deployment of
fundamental network functions (including forwarding, session management, and policy
management) without regard to location, in line with the current trends of network function
modularization and cloudification. With a shared context and a modular design, the per-user
network will do away with message exchange between conventional network functions.

To realize the user-centric 6G network, architectural redesign of the core network is required,
following the paradigm shift from “Network Function-focus” to “user-focus”, allowing users to
participate in network service creation and operation, while also giving users full control over
data ownership. To achieve this design, the network architecture is envisioned to be separated
into user service nodes (USN) and network service nodes (NSN), which will be adaptable for
activities, such as collaborative sensing and distributed learning in order to spread AI

Dissemination level: Public | 150

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

applications on a broad scale across the edge-cloud continuum, in accordance with the
EUCloudEdgeIoT [EoCIoT] European initiative.

9.4 REFERENCES
[Fer23] S. Fernandez, M. Montagud, D. Rincón, J. Moragues, G. Cernigliaro, “Addressing scalability for real-
time multiuser holo-portation: Introducing and assessing a multipoint control unit (MCU) for volumetric
video”, ACM Multimedia’23, October 2023
[Lop24] A. López, A. AbdelNabi, D. Camps-Mur, M. Catalan-Cid, M. Montagud, “NetXRate: O-RAN enabled
network assisted rate control for XR services”, IEEE GLOBECOM’24, December 2024
[Mon24] M. Montagud, et al., "AwareXR: A NaaS architecture to enhance XR services over beyond 5G
networks", IEEE Network, To Appear in 2024
[Yer24] I. Yeregui, D. Mejias, G. Pacho, R. Viola, J. Astorga, and M. Montagud, “Edge Rendering Architecture
for multiuser XR Experiences and E2E Performance Assessment”, IEEE BMSB’24, June 2024
[Tso24] Tsolkas, D., Merino, P., & Dieudonne, M. (2024). THIRD PARTY EXPERIMENTATION: Engagement
process and technical information. Zenodo. https://doi.org/10.5281/zenodo.13594165
[SoftNet24] Tsolkas, Dimitris, David Artuñedo Guillen, Anastasius Gavras, Christos Tranoris, Sándor Laki,
Antonio Skarmeta Gómez, João Paulo Barraca, George Makropoulos, e Ricard Vilalta. «Network & Service
Management Advancements - Key Frameworks and Interfaces Towards Open, Intelligent and Reliable 6G
Networks». Zenodo, 28 dicembre 2024. https://doi.org/10.5281/zenodo.14234898.
[EdgeCloud] CAMARA Edge Cloud https://github.com/camaraproject/EdgeCloud/
[ITU-T-Y3090] ITU-T “Digital twin network – Requirements and architecture”, 2023
[EoCIoT] The EUCloudEdgeIoT.eu initiative https://eucloudedgeiot.eu/

Dissemination level: Public | 151

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

10 6G ARCHITECTURAL DEFINITION
In the following we analyse and discuss the major architectural trends and opportunities that
emerge towards the upcoming standardization work on 6G, starting from the transitions from the
5G technology. To allow for a smooth and faster introduction of 6G services, the main option for
migration between 5G and 6G is to use a so called “evolved 5GC”, where 6G can reuse existing 5GC
NFs if possible and introducing new dedicated 6G NFs to support new 6G functionality, see section
2.3 for more details. In addition, for interworking between 5G and 6G, Multi-Radio Spectrum
Sharing (MRSS) is seen as the main option, see Figure 2.2.

10.1 MODULAR ARCHITECTURE DESIGN

10.1.1 MULTI LAYERED ARCHITECTURES FOR NTN INTEGRATION

A core objective of future 6G networks is truly global, resilient coverage. This will be achieved
by natively integrating Non-Terrestrial Networks (NTNs) with terrestrial infrastructure, creating
a multi-layered 3D architecture. As discussed in Section 4.1, This architecture consists of
satellites, High-Altitude Platforms (HAPs), airborne nodes, and traditional ground-based cells.
This discussion explores the vision, challenges, and enabling technologies behind these multi-
layered NTNs, emphasizing how converging orbital, aerial, and terrestrial segments can deliver
ubiquitous connectivity.

The concept of a 3D multi-layered NTN architecture unifies terrestrial networks with various
non-terrestrial nodes, such as satellites and HAPs at different altitudes. This multi-layered
configuration expands coverage to remote and sparsely populated regions, including maritime
areas, and provides backup and resiliency for critical communications. Key design considerations
include multi-orbit integration, combining Geostationary (GSO) and Non-Geostationary Orbit
(NGSO) satellites (LEO, MEO) for both wide-area broadcast services and low-latency links.
Flexible airborne nodes, like HAPs, specialized drones, or other aerial platforms, can be deployed
opportunistically to manage capacity surges or event-based connectivity, rather than establishing
permanent global coverage layers. Unified radio and access are expected, with native
convergence at the radio interface, enabling handsets and IoT devices to seamlessly connect to
terrestrial gNBs or non-terrestrial nodes. By combining multiple layers, each optimized for
different coverage and performance targets, 6G networks can dynamically adapt resources to
localized needs while maintaining wide-reaching coverage.

Multi-layered architectures inherently enhance resiliency by offering failover paths and

diverse connectivity options. In emergencies or disasters where terrestrial networks fail,

Dissemination level: Public | 152

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

satellites or HAPs can maintain at least minimal service, providing backup infrastructure. Store
and forward mechanisms, using periodic or intermittent satellite links, enable local data storage
and forwarding upon re-establishing ground connectivity, which is vital for delay-tolerant IoT
applications like agriculture, livestock tracking, and maritime operations. Multi-layer
redundancy, using multi-orbit constellations (GEO and LEO), ensures that if one link is
compromised due to congestion or poor channel conditions, other layers can carry essential
traffic.

To accommodate seamless integration, 3GPP is developing standards to support non-

terrestrial access within the 6G system. Distributed 5G/6G core functionalities (AMF, SMF, UPF)
can be partially hosted in space to reduce latency and manage intermittent feeder links. A split-
core architecture, with both on-ground and on-board network functions, allows flexible
deployment, especially for low-density satellite constellations. A hierarchical MANO framework
will cover ground, aerial, and space infrastructure. Domain-level orchestrators manage local
resources, while a global-level orchestrator coordinates cross-domain and mobility management,
ensuring smooth handovers as nodes traverse orbits or airspace. Extending SDN principles to the
user plane enables fine-grained forwarding control across multiple domains and network
segments, simplifying end-to-end provisioning of data paths, including satellite feeder links and
terrestrial backhaul.

This unified 3D architecture positions 6G to serve diverse use cases. It enables direct
smartphone connectivity, extending 6G coverage to remote areas with a consistent user
experience, potentially including limited indoor connectivity. For delay-tolerant IoT, large revisit
times of low-density LEO constellations and store-and-forward mechanisms support agriculture
monitoring, livestock management, maritime navigation, and asset tracking. Flexible broadband
connectivity through ultra-small aperture terminals, suitable for in-vehicle, drone-mounted, or
airborne platforms, is provided for vehicle and drone broadband. Finally, it supports safety-
critical operations by connecting aviation and space control systems, ensuring uninterrupted
data exchange even when conventional terrestrial links are unavailable.

10.1.2 TIME CRITICAL AND DETERMINISTIC NETWORKING

INTEGRATION

Emerging 6G applications, such as extended reality (XR), smart farming, and adaptive
manufacturing, require end-to-end time-critical communications with stringent latency,
reliability, and deterministic performance. Achieving these goals in inherently stochastic
wireless systems, especially when also integrating compute elements (like edge computing) and
deterministic networking technologies (like TSN and DetNet) is a significant challenge.

Dissemination level: Public | 153

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Future 6G networks must allow to incorporate stochastic components, accepting that latency
variation, packet loss, and jitter are inevitable in wireless and distributed compute environments.
These networks must also characterize and predict these stochastic behaviours, such as latency
distributions and reliability levels, to enable proactive resource planning and management,
mitigating variances using mechanisms like packet delay correction or buffering to offset jitter
and keep time-critical packets within acceptable bounds.

Dependable time-critical services must account for both network and computational latency
introduced by edge or cloud processing. Time-aware edge computing, embedding time-
awareness into compute infrastructure (like time-synchronized edge nodes), ensures tasks are
scheduled and completed within predictable deadlines. This requires that end-to-end feedback
loops are established, considering the entire chain from sensors and controllers to actuators,
allow the network to dynamically adjust resource allocations and QoS parameters in real time,
adapting to latency variations within any sub-component and tune application-level components
according to the system load or changes in network characteristic.

To unify deterministic and non-deterministic domains, a horizontal 6G architecture should

expose network capabilities to higher-layer applications through a well-defined service interface.
Performance monitoring and prediction, using continuous monitoring of KPIs (latency, packet
delay variation, reliability), enables data-driven forecasting of whether the network can sustain
the requested QoS. Time-aware configuration allows network control entities to orchestrate
resources, enforcing tight latency bounds across TSN/DetNet and wireless sections, creating
consistent performance "islands" end to end.

Future 6G systems will integrate TSN/DetNet-based deterministic networks with wireless

stochastic segments and compute nodes. This requires holistic traffic handling, where TSN or
DetNet understands each sub-component's latency characteristics, including variable wireless
links, edge compute tasks, and buffering points. Scalable architectures are necessary, as
algorithms and protocols must handle large-scale deployments of sensors, machines, and devices.
Architectural components should plan resources based on probabilistic performance while
meeting application-specific reliability targets. End-to-end orchestration, through a unified
management framework, ensures cross-domain configuration, monitoring, and adaptation at
scale, bridging traditional deterministic subnetworks with next-generation, AI-driven wireless
domains.

As discussed in this white paper "deterministic networking" will transition into a dependable
time-critical communications paradigm, where network performance becomes predictable and
can be matched to the application needs. This is maintained through time-awareness, network
performance observability, adaptive management, and security-by-design. Time-awareness

Dissemination level: Public | 154

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

ensures consistent, synchronized operations across network and compute layers. Adaptive
management allows fine-tuned responses to real-world variations in traffic, mobility, and
compute load. Security-by-design recognizes the critical nature of time synchronization and
performance assurances. By orchestrating communications and computations holistically, 6G
will deliver next-level support for demanding verticals, reliably connecting sensors, controllers,
and actuators across diverse domains and advancing a new era of networked intelligence.

10.1.3 INTEGRATION OF SENSING AND DIGITAL TWINNING

Building on the vision of Integrated Sensing and Communication (ISAC), future 6G networks
will incorporate distributed sensing architectures spanning heterogeneous devices,
reconfigurable surfaces, and multi-modal data sources. Achieving time-critical and deterministic
performance in such complex environments requires an end-to-end design that accounts for
varying capabilities, semantic-aware processing, and advanced orchestration of communication,
computation, and sensing resources.

A core objective of distributed sensing is to collect, fuse, and exploit data from multiple
heterogeneous devices (Sensing Receiver Nodes, SRNs) to track both passive and active targets
over large areas. To optimize distributed sensing under the high dimensionality of multi-modal
data, a semantic plane is introduced. Its key functionalities include context extraction,
interpreting and managing the “meaning” of data rather than raw bits, reducing overhead and
improving relevance. It also includes dynamic adaptation, aligning sensing, communication, and
computation tasks with specific system goals (e.g., continuous target tracking, minimal latency).
Semantic modules and interfaces ensure that all nodes cooperate under common semantic goals,
enabling flexible data sharing and efficient resource utilization.

Future networks will expose sensing-as-a-service capabilities to internal network functions

and third-party applications. A dedicated Sensing Management Function (SeMF), or an extension
of existing location services, will coordinate sensing procedures, manage data flows, and enforce
privacy and security. Secure interfaces for collecting, processing, and distributing sensing data
will accommodate dynamic trust levels and avoid network overload.

A way of achieving precise network configuration without overloading the network is to

introduce AI-Driven Network Digital Twins (NDTs) which offer a virtualized replica of the
physical network and enable predictive optimization. Closed-loop management, using
simulations running in real time (online NDT) or offline “what-if” scenarios, will inform dynamic
resource reallocation or reconfiguration. Federated simulation, integrating multiple domain-
specific simulators (RAN, optical, compute, etc.), will allow large-scale scenario testing. MLOps
principles will streamline the design, training, and deployment of AI models across network

Dissemination level: Public | 155

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

domains, ensuring consistent performance monitoring and retraining based on real

measurements.

10.1.4 SUSTAINABILITY MANAGEMENT

Enforcing sustainability in the network operation has to be achieved with specific architectural
components, such as a Sustainability Monitoring Plane (SMP) designed to manage different
sustainability need coming from telecommunications networks, industry verticals, end users, and
associated services in a 6G networks. This includes comprehensively addressing six dimensions
of sustainability: environmental (resource efficiency, energy consumption), economic (cost-
effectiveness, profitability), societal (user accessibility, inclusivity), technological (innovation,
reliability), regulatory (compliance, standards adherence), and ethical (privacy, transparency).
This generalized monitoring plane is similar to well-known control and data planes, and will
continuously gather and analyse data from multiple sources, such as industrial sectors, consumer
activities, network operators, and smart grid operators.

Figure 10.1 SMP overarching networks, verticals, end user, and energy suppliers to holistically
monitor sustainability indicators

As illustrated in Figure 10.1, the SMP will establish bidirectional communication with multiple
stakeholders—including network operators, vertical industries, end users, and energy
suppliers—to systematically collect and evaluate sustainability indicators spanning
environmental, societal, and economic dimensions. Hence, the work on the definition of the SMP
shall work on:

• Functionalities and Interfaces: SMP capabilities and its integration with diverse systems
and domains.

Dissemination level: Public | 156

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

• Data Exchange: Relevant information flows among stakeholders to effectively monitor

sustainability indicators at different temporal scales.
• Sustainability Metrics: Criteria for assessing and evaluating the various sustainability
dimensions.
• Data Management Policies: Guidelines governing data sharing and handling practices
aimed at enhancing overall sustainability efforts.

10.1.5 GLOBAL SERVICE BASED ARCHITECTURE

The application of the SBA architecture opens the way for the possibility of promoting it
towards other domains in the network. The capability of efficiently promote the consumer
producer paradigm in a data-driven manner makes this technology a good candidate for its
inclusion in other domains beyond Core, as in 5G. In particular, integrating the GSBA concept in
the RAN has the potential to improve several aspects of the radio access network.

• Improved Scalability: SBA decouple network functions into modular services, allowing
networks to scale dynamically based on demand. This flexibility is essential for handling
traffic spikes or expanding capacity without significant hardware investments.
• Enhanced Flexibility and Modularity: The modular design of SBA enables independent
development, deployment, and management of network functions. This approach
supports agile updates and innovation without disrupting the entire network.
• Improved Automation and Orchestration: SBA supports advanced automation tools
and orchestration frameworks. By using programmable interfaces and machine-readable
APIs, networks can automate tasks like resource allocation, fault detection, and recovery.
• Better Resource Management: Fine-grained control over individual services enables
better monitoring and allocation of resources.
• Resilience and Reliability: SBA supports fault-tolerant designs where failures in one
service do not cascade across the network. This architecture improves the overall
reliability and uptime of the network.
• Future-Proofing: With its modular and API-driven approach, SBA is well-suited to adapt
to evolving standards, protocols, and technologies, ensuring long-term relevance and
reduced need for overhauls.

Dissemination level: Public | 157

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Figure 10.2 A possible architecture for the GSBA integration in the RAN

10.2 CLOUD CONTINUUM MANAGEMENT

The 6G network architecture will be very integrated, glueing together advanced connectivity
and computational power from the most deep-edge sub-networks, like those found in vehicles or
robots, through edge sites, and all the way to cloud infrastructure. This integration allows for a
seamless flow of data and processing, enabling tasks like autonomous driving or robotic control
to be executed with optimal efficiency.

To achieve this, the proposals listed in this white paper suggest the utilization of a resource
pool across this continuum, where tasks can be dynamically offloaded to the most suitable
location and key to this vision are new architectural enablers. The Communication & Computing
Resources Exposure Function (CCREF) extends network exposure to provide real-time
awareness of diverse resources, including computing, connectivity, and AI, across the entire 6G
network. The Communication & Computational Resources Management Function (CCRMF)
orchestrates advanced resource-sharing policies, dynamically allocating capacity to meet
performance demands. Network Intelligence Functions (NIF) leverage AI and machine learning
for proactive network management, enhancing existing analytics. The introduction of a Compute
Continuum Layer (CCL) abstracts heterogeneous computing elements, ensuring seamless access
to optimal resources for network functions and applications, leading to performance and energy
efficiency gains. Additionally, Trusted Execution Environments (TEEs) provides a hardware-
based secured platform for confidential computing, reinforcing integrity and confidentiality of
data as it moves across the network continuum.

Besides these novel components, multi-access Edge Computing (MEC) is enhanced by defining
application slices that incorporate both network and computing requirements, working in
tandem with network slicing to fulfill diverse QoS and compute demands. AI/ML-driven
orchestration enables real-time data analytics, proactive resource scaling, automated healing,

Dissemination level: Public | 158

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

and threat detection, with distributed and federated learning ensuring privacy and efficiency.
Zero-touch closed loops support self-configuration and optimization, while hierarchical
orchestrators coordinate across administrative boundaries, allowing all devices, from resource-
constrained IoT to cloud sites, to participate in collaborative processing. In essence, this 6G
management approach unites deep edge, edge, and cloud resources into a unified framework,
facilitating the dynamic and efficient allocation of connectivity and computing resources, thus
ensuring high performance, flexibility, and sustainability for a wide array of services.

10.3 INTEROPERABILITY AND GLOBAL OPERATION

6G architecture should enable zero-trust principles to enforce how infrastructure capabilities
are shared and consumed across diverse stakeholders and domains. By integrating a Zero-Trust
Layer (ZTL) into the architecture, the network achieves granular security, continuous evaluation,
advanced analytics, and adaptable business models, moving beyond traditional perimeter-based
trust systems.

This approach fosters cooperative control between network operators and service providers,
mirroring hyperscale cloud operational models. Instead of requiring complete trust, participants
share only essential performance and analytics data through managed interfaces. This ensures
secure feedback loops, feeding service provider data into the Network Data Analytics Function
(NWDAF) for personalized network optimization without compromising confidential business
information. This cooperative loop enables continuous optimization, automating resource
adjustments to meet each provider's unique quality of experience (QoE) metrics. The ZTL
framework emphasizes both vertical and horizontal exposure, catering to a wide range of use
cases.

In this context, vertical exposure integrates feedback from service providers into AI-driven
network analytics and management, allowing providers to adjust network functions, slices, or
other configurations to align with their application-level metrics. Horizontal exposure facilitates
global operation in multi-operator scenarios, such as international roaming, enabling direct,
secure interaction among different network operators and service providers without relying on
vulnerable trust-based models.

A decentralized identity model replaces legacy roaming agreements, enabling visited

operators to directly charge global end-users while providing full visibility to the home operator.
This eliminates inefficient data routing, reducing latency and transport costs, and fosters new
business relationships among operators, hyperscalers, and vertical providers. By decoupling user
identity, integrating AI/ML-driven analytics, and utilizing distributed ledgers, this approach
paves the way for real-time, zero-trust operations in future communications networks.

Dissemination level: Public | 159

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Besides, in line with Zero Trust principles, decentralized identity management can be
integrated within an authentication framework that operates seamlessly across multiple
proprietary domains. Proposed framework leverages attribute derivation techniques. Based on
the self-sovereign identity paradigm, attributes ensures that only the minimal and essential
information required to validate access to network resources and services is disclosed. It
safeguards customer identity, mitigating traceability and linkability while supporting privacy-
preserving operations. From an operator’s perspective, the architecture incorporates dedicated
issuers and verifiers to manage part of the credential lifecycle. Issuers generate and
cryptographically sign verifiable credentials containing the derived attributes, thereby providing
a secure proof of identity and access rights. Verifiers, strategically deployed across the network,
authenticate these credentials in real time, ensuring that only users with valid credentials can
access network services.

This new wave of authentication schemes is the principal enabler for operator-agnostic access,
allowing users to select and access services from any operator based on the current network state
and service requirements. Additionally, the framework marks a significant advancement for
proximity services, delivering fast and privacy-preserving authentication that supports highly
dynamic service compositions driven by users’ locations.

10.4 AI DRIVEN NETWORK MANAGEMENT AND

ORCHESTRATION

With the advent of 6G, network management and AI-driven orchestration are key components
for accomplishing autonomous and dynamic network operations. In this whitepaper we outline
the principal role of AI to enable, Intent-based closed-loop management, where ongoing
monitoring, analysis, decision-making, and execution are integrated within an orchestration
framework

In presented approaches, AI is not only responsible for automation but also allows for
comprehensive operational management of network functions and resources distributed across
the cloud continuum (Far-edge, RAN, Edge, transport, Core, etc.). A critical function is the
integration with advanced AI/ML frameworks, comprising MLOps for continuous model
deployment and retraining, and DataOps for robust, high-quality data pipelines, which work
together to ensure that real-time telemetry is accurately processed and actionable insights are
derived. Key to the concept is the use of intent-based management. High-level service intents,
reflecting user requirements and network policies, are translated through standardized APIs into
detailed network and service configurations. This process is supported by a distributed, multi-

Dissemination level: Public | 160

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

agent orchestration model that coordinates the control of both intra- and inter-domain resources.
The architecture supports flexible closed-loop mechanisms, allowing its features to be expanded
as new technologies or tools are developed, as well as proactive fault management, domain
optimization, resource reassignment, and service migration, which are capabilities necessary to
adapt the network to the continuous changes it faces.

Furthermore, the AI-driven orchestration framework natively incorporates elements of

trustworthiness. Starting from intent modelling, involving the five dimensions of trustworthiness
to express requirements. Passing through explainable AI, which is embedded in the system to
provide transparency in the decision-making process, enabling oversight by management teams.
In addition to privacy methods implemented to ensure that sensitive data remains secure
throughout the operation’s lifecycle, especially useful in federation scenarios. Moreover,
integration of digital twin technologies enables real-time simulation and reliable predictions of
KPIs; these digital twins allow for the evaluation of measures and countermeasures before they
are applied, as well as enabling the training of ML agents at an unprecedented speed.

10.5 REFERENCES
[SUST6G] SUSTAIN-6G project “SUSTainability Advanced and Innovative Networking with 6G”, 2025.
Available: https://sustain-6g.eu/

Dissemination level: Public | 161

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

11 ABBREVIATIONS AND ACRONYMS

Acronym Description

3GPP 3rd Generation Partnership Project

4G Fourth Generation Mobile Network

5G Fifth Generation Mobile Network

5GC 5G Core

5GCN 5G Core Network

5GS 5G System

6G Sixth Generation Mobile Network

AAA Authentication, Authorization, and Accounting

ABAC Attribute-Based Access Control

ADR Adaptive Data Rate

AF Application Function

AI Artificial Intelligence

AICP AI-driven Multi-Stakeholder Inter-Domain Control-Plane

AMF Access and Mobility Management Function

AN Access Network

AOA Angle of Arrival

AOSP Android Open Source Project

AP Access Point

API Application Programming Interface

AR Augmented Reality

ASIC Application-Specific Integrated Circuit

ATSSS Access Traffic Steering, Switching, and Splitting

B5G Beyond 5G

BC Blockchain

BCN Blockchain Network

BMSB Broadcast and Multicast Service Broadcast

Dissemination level: Public | 162

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

BS Base Station

BSM Basic Safety Message

BVT Bandwidth Variable Transceivers

CA Carrier Aggregation

CAM Cooperative Awareness Message

CAMARA Cloud Application Management for Open Networks

CCL Compute Continuum Layer

CCREF Communication & Computing Resources Exposure Function

CCRMF Communication & Computational Resources Management Function

CDN Content Delivery Network

CF Computing Function

CL Cloud Layer

CN Core Network

CNCF Cloud Native Computing Foundation

CNF Cloud-Native Function

CNSM Conference on Network and Service Management

COH Coherent Optical

COTS Commercial Off-The-Shelf

CP Control Plane

CPU Central Processing Unit

CS Communication Service

CSA Cloud Service Architecture

CSMF Communication Service Management Function

CTI Cyber Threat Intelligence

CU Centralized Unit

DAR Data Analysis and Reporting

DASH Dynamic Adaptive Streaming over HTTP

DC Data Center

Dissemination level: Public | 163

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

DD Data Distribution

DFT Discrete Fourier Transform

DI Data Integration

DL Downlink

DLT Distributed Ledger Technology

DMMF Domain Mobility Management Function

DMOC Domain Management & Orchestration Component

DPU Data Processing Unit

DSCM Digital Subcarrier Multiplexing

DSP Digital Signal Processing

DSS Dynamic Spectrum Sharing

DT Digital Twin

DTD Document Type Definition

DU Distributed Unit

DUG Data Unit Groups

DWDM Dense Wavelength Division Multiplexing

E2E End-to-End

E2SM E2 Service Model

EC Edge Computing

ECU Electronic Control Unit

EMOC E2E Management & Orchestration Component

EN Edge Node

EPC Evolved Packet Core

ETSI European Telecommunications Standards Institute

EUC End User Computing

FAV Fully Autonomous Vehicles

FC Forwarding Controller

FG Focus Group

Dissemination level: Public | 164

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

FGCS Future Generation Computer Systems

FIPS Federal Information Processing Standards

FL Federated Learning

FLEX Flexible Networks

FM Fault Management

FPGA Field Programmable Gate Array

FR1 Frequency Range 1

FR2 Frequency Range 2

FSO Free-Space Optics

FTM Fine Timing Measurement

FTTH Fiber to the Home

FWA Fixed Wireless Access

GDPR General Data Protection Regulation

GEO Geostationary Orbit

GH Green Hydrogen

GIS Geographic Information System

GMMF Global Mobility Management Function

GND Ground

GNSS Global Navigation Satellite System

GPU Graphics Processing Unit

GSBA Global Service-Based Architecture

GSMA GSM Association

GSO Geostationary Satellite Orbit

GW Gateway

HAL Hardware Abstraction Layer

HAP High Altitude Platform

HC Higher Capabilities

HLS High-Level Synthesis

Dissemination level: Public | 165

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

HMP Hybrid Multiplexing Processing

HPEC High-Performance Embedded Computing

IAB Integrated Access and Backhaul

IBI Information-Based Interface

IBN Intent-Based Networking

IBTM intent-based threat mitigation module

ICAS Integrated Communication and Sensing

ICT Information and Communication Technology

IEC International Electrotechnical Commission

IEEE Institute of Electrical and Electronics Engineers

IETF Internet Engineering Task Force

IML Interactive Machine Learning

IMT International Mobile Telecommunications

IN Intelligent Network

IOTM Internet of Things Management

IP Internet Protocol

IPU Intelligence Processing Unit

IR Infrared

IRTF Internet Research Task Force

ISAC Integrated Sensing and Communication

ISL Inter-Satellite Link

ISO International Organization for Standardization

IT Information Technology

ITU-T International Telecommunication Union - Telecommunication Standardization Sector

JSON JavaScript Object Notation

KPI Key Performance Indicator

LAN Local Area Network

LBS Location-Based Services

Dissemination level: Public | 166

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

LIDAR Light Detection and Ranging

LLS Lower Layer Split

LPWAN Low-Power Wide-Area Network

MAC Medium Access Control

MANO Management and Orchestration

MEC Multi-access Edge Computing

MIMO Multiple-Input Multiple-Output

ML Machine Learning

MLOps Machine Learning Operations

MME Mobility Management Entity

MNO Mobile Network Operator

MRF Media Resource Function

MRSS Multi-Radio Spectrum Sharing

NAS Non-Access Stratum

NF Network Function

NFV Network Function Virtualization

NIC Network Interface Card

NR New Radio

NSA Non-Standalone

NSMF Network Slice Management Function

NTN Non-Terrestrial Networks

OAM Operations, Administration, and Maintenance

O-RAN Open Radio Access Network

OSS Operations Support System

OTFS Orthogonal Time Frequency Space

PDU Protocol Data Unit

PHY Physical Layer

PKI Public Key Infrastructure

Dissemination level: Public | 167

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

PON Passive Optical Network

PPDR Public Protection and Disaster Relief

QoE Quality of Experience

QoS Quality of Service

RAN Radio Access Network

RAT Radio Access Technology

RF Radio Frequency

RIC RAN Intelligent Controller

RLC Radio Link Control

RRM Radio Resource Management

SBA Service-Based Architecture

SDN Software-Defined Networking

SIM Subscriber Identity Module

SLA Service Level Agreement

SMF Session Management Function

SON Self-Organizing Network

SMP Sustainability Management Plane

SRv6 Segment Routing over IPv6

TCO Total Cost of Ownership

TN Terrestrial Network

TPM Trusted Platform Module

TSN Time-Sensitive Networking

UE User Equipment

ULP Ultra-Low Power

UPF User Plane Function

URLLC Ultra-Reliable Low Latency Communication

V2X Vehicle-to-Everything

VIM Virtual Infrastructure Manager

Dissemination level: Public | 168

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

VM Virtual Machine

VN Virtual Network

VR Virtual Reality

WAN Wide Area Network

WDM Wavelength Division Multiplexing

XAI Explainable Artificial Intelligence

XR Extended Reality

ZTA Zero Trust Architecture

Dissemination level: Public | 169

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

12 LIST OF EDITORS & REVIEWERS

Name Company / Institute / University Country

Document editors

Marco Gramaglia Universidad Carlos III de Madrid Spain

Ömer Bulakci Nokia Germany

Xi Li NEC Laboratories Europe Germany

Anastasius Gavras Eurescom GmbH Germany

Chapter 2 editors

Mårten Ericson Ericsson Sweden

Sylvaine Kerboeuf Nokia Bell Labs France

Chapter 3 editors

Marco Gramaglia Universidad Carlos III de Madrid Spain

David Larrabeiti Universidad Carlos III de Madrid Spain

Chapter 4 editors

Mir Ghoraishi Gigays UK

Agapi Mesodiakaki Aristotle University of Thessaloniki Greece

Chapter 5 editors

Mårten Ericson Ericsson Sweden

Anastasius Gavras Eurescom GmbH Germany

Harilaos G. Koumaras NCSR Demokritos Greece

Chapter 6 editors

Xi Li NEC Laboratories Europe Germany

Sebastian Robitzsch InterDigital UK

Chapter 7 editors

Rodrigo Asensio Garriga University of Murcia Spain

Antonio Skarmeta University of Murcia Spain

Harilaos G. Koumaras NCSR Demokritos Greece

Chapter 8 editors

Dissemination level: Public | 170

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Anastasius Gavras Eurescom GmbH Germany

Agapi Mesodiakaki Aristotle University of Thessaloniki Greece

Chapter 9 editors

Dimitris Tsolkas Fogus Innovations & services Greece

Chapter 10 editor

Marco Gramaglia Universidad Carlos III de Madrid Spain

Name Company / Institute / University Country

External Reviewer

Nazli Güney Turk Telekom Türkiye

Dissemination level: Public | 171

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

13 LIST OF CONTRIBUTORS

Name Company / Institute / University Country

Adam Zahir Universidad Carlos III de Madrid Spain

Adriana Fernández- i2CAT Foundation Spain

Fernández

Agapi Mesodiakaki Aristotle University of Thessaloniki Greece

Alessandro Guidotti CNIT Italy

Alessandro Vanelli-Coralli CNIT / University of Bologna Italy

Alexandros Kostopoulos Hellenic Telecommunications Organization (OTE) Greece

Group

Almudena Diaz Zayas University of Malaga Spain

Amr AbdelNabi i2CAT Foundation Spain

Anastasios Zafeiropoulos Institute of Communication and Computer Systems Greece

Anastasius Gavras Eurescom GmbH Germany

Andra Lutu Telefonica Spain

Andres Saavedra NEC Laboratories Europe Germany

Anna Tzanakaki UOA Greece

Anna Umbert Universitat Politecnica de Catalunya Spain

Antonio de la Oliva Universidad Carlos III de Madrid Spain

Antonio Skarmeta University of Murcia Spain

Aurora Ramos Capgemini Spain

Baldomero Coll-Perales Universidad Miguel Hernandez de Elche Spain

Benjamin Barth German Aerospace Center (DLR) Germany

Carlos J. Bernardos Universidad Carlos III de Madrid Spain

Christoph Schmelz Nokia Germany

Christoph Sommer TU Dresden Germany

Christos Verikoukis Industrial Systems Institute Greece

Chrysa Papagianni University of Amsterdam The

Netherlands

Dissemination level: Public | 172

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Dan Marom HUJI Israel

Daniel Camps-Mur i2CAT Foundation Spain

David Larrabeiti Universidad Carlos III de Madrid Spain

David Rico Menendez Universidad Carlos III de Madrid Spain

Diego San Cristobal Ericsson Spain

Dimitris Pliatsios University of Western Macedonia Greece

Dimitris Tsolkas Fogus Innovations & services Greece

Elli Kartsakli Barcelona Supercomputing Center Spain

Emilio Calvanese Strinati CEA France

Fabrizio Granelli CNIT Italy

Fernando Agraz Universitat Politecnica de Catalunya Spain

Filippo Cugini CNIT Italy

George Alexandropoulos NKUA Greece

George Xilouria Space Hellas Greece

Georgios Gardikis Space Hellas Greece

Gergely Pongracz Ericsson Research Hungary

Giada Landi Nextworks Italy

Gilberto Berardinelli Aalborg University Denmark

Giyyarpuram Orange France

Madhusudan

Gourav Prateek Sharma KTH Royal Institute of Technology Sweden

Harilaos G. Koumaras NCSR Demokritos Greece

Henk Wymeersch Chalmers University Sweden

Ignacio Labrador Pavon ATOS Spain

Ioannis Chochliouros Hellenic Telecommunications Organization (OTE) Greece

Group

Ioannis Tomkos UPAT Greece

Ion Turcanu Luxembourg Institute of Science and Technology Luxembourg

Irene Vila Universitat Politecnica de Catalunya Spain

Dissemination level: Public | 173

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

James Gross KTH Royal Institute of Technology Sweden

Jarno Pinola VTT Finland

Javier Garcia Rodrigo Telefonica Spain

Javier Gozalvez Universidad Miguel Hernandez de Elche Spain

Joachim Sachs Ericsson Research Sweden

John Cosmas Brunel University UK

Jordi Perez-Romero Universitat Politecnica de Catalunya Spain

Jose A. Ayla-Romero NEC Laboratories Europe Germany

Jose M. Alcaraz Calero University of the West of Scotland UK

Josep Martrat Eviden Spain

Josep Xavier Salvat NEC Laboratories Europe Germany

Lozano

Joai Fernandes OneSource Consultoria Informática Portugal

Juan Sanchez-Gonzalez Universitat Politecnica de Catalunya Spain

Janos Harmatos Ericsson Research Sweden

Kyriakos Vlachos CNIT Italy

Kostas Ramantas Iquadrat Spain

Lanfranco Zanzi NEC Laboratories Europe Germany

Lechoslaw Tomaszewski Orange Polska Poland

Luis Cordeiro OneSource Consultoria Informática Portugal

Luis M. Pessoa INESCTEC Portugal

Luis Velasco Universitat Politecnica de Catalunya Spain

Marco Fiore IMDEA Networks Institute Spain

Marco Gramaglia Universidad Carlos III de Madrid Spain

Marco Quagliotti Telecom Italia Italy

Maria Christopoulou Space Hellas Greece

Marilet De Andrade Ericsson Research Sweden

Jardim

Mario Montagud i2CAT Foundation & University of Valencia Spain

Dissemination level: Public | 174

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Mark Angoustures Solidshield France

Maxime Compastie i2CAT Spain

Merve Saimler Ericsson Research Türkiye

Miguel Camelo Imec Belgium

Miguel Catalan-Cid i2CAT Foundation Spain

Mir Ghoraishi Gigays UK

Marten Ericson Ericsson Research Sweden

Nicolas Chuberre Thales Alenia Space France

Nikolaos Nomikos National and Kapodistrian University of Athens Greece

Oriol Sallent Universitat Politecnica de Catalunya Spain

Oscar Gonzalez de Dios Telefonica Spain

Ömer Bulakci Nokia Germany

Özgür Akgül Nokia Finland

Pablo Picazo Martinez Universidad Carlos III de Madrid Spain

Panagiotis Botsinis Apple Germany

Panagiotis Gkonis National and Kapodistrian University of Athens Greece

Panagiotis Trakadas National and Kapodistrian University of Athens Greece

Paolo Di Lorenzo Sapienza University of Rome Italy

Pietro G. Giardina Nextworks Italy

Placido Mursia NEC Laboratories Europe Germany

Qi Wang University of the West of Scotland UK

Ramon Casellas CTTC Spain

Ramoni Adeogun Aalborg University Denmark

Robert Kolakowski Orange Polska Poland

Roberto Viola Vicomtech Spain

Rodrigo Asensio Garriga University of Murcia Spain

Ryan Husbands British Telecom UK

Salvatore Spadaro Universitat Politecnica de Catalunya Spain

Dissemination level: Public | 175

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

Sameh Eldessoki Apple Germany

Sandor Laki ELTE Hungary

Sandro Scalise German Aerospace Center (DLR) Germany

Sean Aherne DELL Ireland

Sebastian Robitzsch Interdigital UK

Sokratis Barmpounakis WINGS Greece

Stefan Wänstedt Ericsson Sweden

Sylvaine Kerboeuf Nokia France France

Tarik Taleb ICT-FI Finland

Tomaso de Cola German Aerospace Center (DLR) Germany

Vasilis Tsekenis WINGS Greece

Vincent Lefebvre Solidshield France

Xavi Masip-Bruin Universitat Politecnica de Catalunya Spain

Xi Li NEC Laboratories Europe Germany

Dissemination level: Public | 176

6G Architecture WG Towards 6G Architecture: Key Concepts, Challenges, and Building Blocks

SUPPORTING PROJECTS

Dissemination level: Public | 177

 The SNS JU projects have received funding from the Smart
Networks and Services Joint Undertaking (SNS JU) under the
European Union's Horizon Europe research and innovation
programme.

Website: https://smart-networks.europa.eu/sns-ju-working-groups/