Hexa-X laying the foundation for 6G
The post was updated Jun 13 with links to material from the EUCNC & 6G Summit 2023.
The Hexa-X European flagship research project for 6G is nearing completion after running since January 2021. The vision and aspirations mentioned at the start have served as the guiding principles for the Hexa-X research efforts that have now laid a solid foundation for the development of 6G technology.
Significant outcomes of the project are a comprehensive list of possible use cases, associated performance metrics and societal value indicators, and a set of potential technology enablers that have been subject to proof-of-concept assessments.
The outcomes have been detailed in the final set of deliverables, which were presented at the Hexa-X 6G workshop series and showcased in the demo exhibition at the European Conference on Networks and Communications (EuCNC) and 6G Summit (EuCNC&6G Summit) in Gothenburg, June 6-9, 2023.
In this blog post, Ericsson, as Technical Manager of Hexa-X, provides a summary of the key findings and achievements in Hexa-X so far.
Hexa-X end-to-end architecture overview
The Hexa-X deliverable D1.3 “Targets and requirements for 6G – initial E2E architecture”, presents an overview of the end-to-end 6G architecture consisting of three layers: infrastructure, network service, and application, and highlights key technical enablers.
- The infrastructure layer includes the radio access network (RAN), the core network (CN), and the transport network. This layer provides physical resources to host network service and application layer elements. It has RAN enhancements for extreme throughput, low latency, and high reliability and availability. The 6G architecture integrates various (sub)network solutions into a network of networks that can adapt to new topologies with the flexibility to meet the requirements of both extreme performance and global service coverage, surpassing the capabilities of 5G. The infrastructure layer for 6G is envisioned to accommodate new enablers such as localization and sensing to enable services beyond communication.
- The network service layer is envisioned to be entirely cloud-based, with microservices used for all network functions. This approach will lead to an intelligent and efficient architecture, incorporating AI, providing AI as a service (AIaaS), programmability, and network automation. A cloud native approach will streamline RAN and CN architectures by removing duplication and complexity. The concepts of edge and far-edge computing become more relevant for the 6G architecture and services. Cloud native technologies will be required to create cloudlets at the edge of the network, with application-to-application and function-to-function communication capable of handling a large number of interconnected assets with flexible mesh topologies. This layer contains an exposure framework and an integration fabric, which creates a communication channel between various domains and facilitates smooth interoperation and networking across diverse domains.
- The application layer is the topmost layer of the Hexa-X E2E architecture which interacts directly with end-user applications and facilitates the exchange of data and information.
The network management and orchestration is moving toward full automation with the adoption of artificial intelligence (AI) and machine learning (ML) technologies. The goal is to provide an orchestration continuum that efficiently supports reliability, flexibility, resilience, and availability. The orchestration continuum concept enables seamless orchestration across device-edge-cloud and addresses changes in infrastructure, requirements, and failures.
Security and privacy mechanisms are crucial for all layers of the 6G architecture. Hexa-X has identified several 6G security enablers, including trust foundations, physical layer security, and quantum security for the infrastructure layer. Various types of cryptography, including quantum-safe cryptography, are also important for secure communication and storage across all layers. Privacy-enhancing technologies are necessary for sensitive data processing in all layers, including the management domain. Secure, privacy-preserving, and trustworthy AI/ML methods are essential for the adoption of AI/ML in the 6G networks. Distributed ledger technologies, such as blockchain, establish trust among participants without any central authorities by distributing the responsibility for maintaining a shared record of transactions. The decentralized approach reduces the risk of fraud and manipulation while increasing transparency and trust in various domains.
Radio access technologies in 6G
Hexa-X has explored various aspects of 6G radio system design to meet the requirements of emerging use cases, for example, a link data rate of 100 Gbps. The project has:
- conducted channel measurements.
- developed channel and hardware models for frequencies above 100 GHz.
- designed radio frequency (RF) transceiver architectures.
- explored deployment scenarios with cellular-based and distributed multiple input multiple output (D-MIMO) transmission approaches.
- studied signal processing topics such as waveform, beam management, and D-MIMO transmission schemes.
A comprehensive performance analysis was also conducted with varying objectives, such as link-level evaluation of modulation under hardware (HW) nonidealities, and system-level analysis to gauge the efficacy of different design and deployment options.
Deliverable D2.3, “Radio models and enabling techniques towards ultra-high data rate links and capacity in 6G”, provides technical guidelines for RF transceiver design to achieve link data rates at a certain range.
In this blog post, we discuss several parameters and degrees of freedom, including bandwidth, signal-to-noise ratio (SNR), analog-to-digital converters (ADCs), HW nonidealities, and waveform impact on the power amplifier (PA) backoff. In addition, we discuss RF implementation aspects and strategies to optimize the system such as channelization and carrier aggregation (CA).
Based on the analysis, it is concluded that using sub-array-based RF transceivers would be a practically feasible architecture for above 100 GHz, which implies analog beam steering per subarray and digital precoding over multiple RF chains. Different configurations for the RF chains are possible, for example, spatial multiplexing, or implementation of ultra-wideband waveforms following a CA approach. However, conventional initial beam access is impractical due to the large number of beams to be probed. As a result, it is necessary to consider using side information for beam management.
Hardware models for local oscillator phase noise and PA nonlinearity have been developed and their accuracy verified based on measurements and circuit analysis. These models are used for the following tasks:
- To design signal processing methods, for example, waveforms and modulation schemes that are robust to hardware impairments like phase noise or PA nonlinearity
- To develop methods for estimation of hardware impairments and for designing techniques to compensate the impact of hardware impairments.
- To carry out performance evaluations of the system subject to hardware impairments.
Performance evaluations are conducted in the presence of one or more hardware impairments, with particular attention given to Discrete Fourier Transform-Spread Orthogonal Frequency Division Multiplexing (DFT-Spread OFDM) and single carrier (SC) schemes due to their greater tolerance to PA nonidealities compared to OFDM. Furthermore, zero-crossing modulation with one-bit quantization is considered a potential candidate for energy-efficient modulation schemes.
At EuCNC & 6G Summit 2023, Hexa-X demonstrated a feasible implementation approach for 6G ultra-wideband at sub-THz, which aims to relax the hardware requirements of the transmitter and receiver, particularly the ADC, while potentially impacting other blocks such as PAs with reduced out of band (OOB) emission due to narrowband operation in individual transmitter/receiver chains. The demonstrated architecture offers considerable flexibility in spectrum management and allows the implementation of wideband across non-contiguous bands.
Hexa-X has developed multiple transmission schemes and analyzed the performance of D-MIMO systems in terms of spectral efficiency, throughput, and coverage, along with techniques to cope with blockage. D-MIMO can be beneficial across a wide range of frequencies, with specific advantages for each frequency range. At lower frequencies, such as sub-6GHz, D-MIMO is deployed primarily to achieve high spectral efficiency. At higher frequencies, such as sub-THz, distributed antenna systems are necessary to overcome the challenges of an unreliable radio channel, extend coverage, and provide reliable communication links to user equipment (UE). The distributed architecture enables the use of serving antennas located closer to UE, which results in a more dependable link. D-MIMO can be implemented with various design options and optimization goals, as depicted in Figure 2.
Watch the recorded waveform demo on Youtube.
Localization and sensing in 6G
Hexa-X envisions the integration of the radio infrastructure in 6G to provide advanced services like positioning, localization, and sensing for the next generation of mobile networks. Building on the localization capabilities of 5G, the localization accuracy of the UEs is expected to be further enhanced, and sensing aspects, such as the detection of passive objects, are gaining significant momentum as an additional promising feature. These advancements are driving the development of a solid foundation for 6G, one that will enable transformative applications and services that extend beyond traditional communication. Sensing refers to the detection of events or changes in an environment, based on radio signals and it can be classified into two types:
- Radar-like sensing involves processing radio signals to extract information about the position, velocity, and state of unconnected objects, and can be used for mapping static objects or tracking moving objects.
- Non-radar-like sensing performs detection and tracking based directly on the received waveform or features extracted from the received waveform and can be used for applications including pollution monitoring, and weather monitoring.
The project has investigated the role of localization and sensing in the 6G ecosystem, and the recently published Deliverable D3.3 “Final models and measurements for localization and sensing” presented the basic sensing processes and functional view, comprising the sensing function layer, the sensing data processing layer, and the emerging application and service layer, as well as aspects related to management and orchestration, and sensor fusion.
Hexa-X has developed a set of 19 performance indicators for localization and sensing. A few of these are considered particularly relevant for use case and gap analysis, including accuracy, latency, availability, and scalability for localization, and accuracy, range/distance resolution, velocity and angle, unambiguous range, latency, and availability for sensing.
Hexa-X outlines the performance indicators and requirements for 17 localization and four sensing use cases. It conducted a baseline evaluation based on 3GPP Release 16 for localization, radar, and lidar for sensing. The analysis revealed that 5G can support some localization use cases under certain conditions, but not most of them, and that conventional sensors may not meet all the sensing requirements. Considering these findings, 6G, as the next generation of mobile communication networks, will need to leverage a wide bandwidth and large antenna arrays, among other key enablers, to fully support and meet the sensing requirements of the identified use cases.
The localization and sensing capabilities are anticipated to facilitate the emergence of novel applications and services, including location-based services that can utilize the absolute or relative location of targets to optimize a particular objective. The required information can be obtained either through UE localization or radar-like sensing without the need for UE. Some examples of such services include:
- searching or finding UE in real time
- historical tracking of certain objects
- geo-fencing for determining whether a UE or an object is within a certain predefined area
- avoiding collisions between two mobile targets or between a mobile target and a static area
- routing and navigation of targets through the environment to optimize the network
Services offered by this functionality could be either used for services outside the network or for the optimization of the network operations. The integration of contextual information, such as the current position of UE and its planned trajectory, could enable internal location-based services to optimize resource allocation.
Watch the JCAS demo on Youtube.
Connecting intelligence for 6G networks
It is anticipated that forthcoming 6G network functionalities and use cases will incorporate a variety of learning and intelligence aspects, including AI-driven air interface design, data management, compute and processing function optimization, network automation, and service availability, among others. They will also require the establishment of reliable mechanisms for secure and trustworthy operations. To tackle the research challenges in connecting intelligence, Hexa-X has studied these areas and in the recently published Deliverable D4.3 “Deliverable D4.3 AI-driven communication & computation co-design: final solutions” to present the developed technical enablers, algorithms, and solutions for AI-driven communication and computation.
AI/ML for RAN performance enhancement: Hexa-X has developed and evaluated methods for enhancing network performance using AI/ML techniques. These methods have been shown to provide considerable benefits, including increased spectral efficiency, improved reliability, and higher energy efficiency. The proposed approaches leverage ML/AI functionalities either at the receive or transmit side to optimize functionalities or use end-to-end learning to jointly optimize both transmitter and receiver side RAN functionalities. Hexa-X has developed:
- an AI-empowered receiver for PA non-linearity compensation, leading to higher spectral efficiency and throughput and enhanced energy efficiency in RAN
- AI-based enhancement for sub-Thz by joint learning of waveform and receiver to mitigate the impact of PA non-linearity, leading to reduced out-of-band emissions and higher bit rates
- AI/ML-based beam selection for D-MIMO systems to reduce beam scanning overhead for D-MIMO systems and improve spectral efficiency
- an AI/ML-based decoder to enhance the energy efficiency of LDPC decoding
- AI/ML-based channel estimation to reduce complexity and improve energy efficiency
AI/ML for orchestration & management: The use of AI/ML solutions has also resulted in performance improvements in the management and orchestration network domain. Examples of these solutions include AI/ML-based predictive orchestration and automated user plane function scaling through distributed AI/ML methods. The EuCNC & 6G Summit 2023 event will feature a demonstration of the benefits of AI/ML solutions in this domain, showcasing the concept of data-driven device-edge-cloud continuum management in a smart city use case scenario.
AI/ML for sustainable 6G: To fully leverage the benefits of AI for performance enhancement functions, it's crucial to consider the complexity and energy consumption of these solutions. Hexa-X has addressed these questions as part of its investigation into 6G network sustainability. Well-designed AI architectures have demonstrated improved energy efficiency without compromising performance. This sustainability perspective is further reinforced by the development of reduced complexity solutions for optimization problems that are typically intractable using conventional methods.
6G as an efficient AI platform: Hexa-X has a vision for 6G networks to serve as an intelligent and reliable platform for AI applications. To achieve this, Hexa-X has proposed design concepts from three different perspectives: supporting network services, challenges of model training, and real-time inference functions. In terms of supporting network services, Hexa-X has suggested open interfaces and data structures that are aligned with the requirements of AI applications. Additionally, distributed training methods, including federated learning, are being investigated to reduce and balance the large data load that could strain the network. For real-time inference functions, Hexa-X is addressing the challenges of distributed operation, high device density, low latency, and high accuracy requirements through joint communication and compute solutions. To ensure the trustworthy operation of AI functions, Hexa-X is also developing enablers for designing AI systems that are transparent, explainable, reliable, and fair, while protecting data privacy and security.
Secure, private, and trusted AI: Hexa-X covers AI trustworthiness in three aspects: security, privacy and explainability, and contributes to these aspects with several technical enablers. The main goal of these enablers is to design and develop AI systems that are transparent, reliable, and fair, and ensure data privacy and security. It is also important for AI systems to be transparent and explainable in decision-making processes and be understandable and trustable by humans. Therefore, developing and deploying trusted AI systems is essential. Hexa-X developed mechanisms to increase the robustness of AI-driven methods against adversarial attacks, privacy-preserving methods based on federated learning, and explainable AI methods.
During EuCNC & 6G Summit 2023, Hexa-X demonstrated a framework for the federated learning of explainable AI models using a testbed comprised of real apps and devices, as well as a real-time network simulator. This demonstration aims to predict the quality of service (QoS) experienced by users while maintaining privacy and transparency, which promotes a trustworthy mobile network.
Watch the recorded Fed-XAI demo on Youtube.
Network evolution and expansion for 6G
Hexa-X has developed architectural enablers to support AI integration, flexible network design, and network programmability for 6G networks. In the recently published Deliverable D5.3 “Final 6G architectural enablers and technological solutions”, these enablers are presented and outlined in Figure 6 below:
Intelligent network enablers: Intelligent network enablers are focused on developing a fully integrated AI and programmable network and are envisioned to exist in both devices and the network itself. These enablers are categorized into frameworks that cover specific functionalities and supporting mechanisms needed to implement the following intelligent networks:
- The AI-as-a-Service (AIaaS) framework facilitates the management, distribution, and training of AI models to AI agents.
- The Federated Learning as-a-Service (FLaaS) framework provides tools to discover and join a federation of UE to utilize federated AI models and possibly participate in collective model training based on distributed local datasets.
- The analytics framework provides basic functionality for data collection, storage, and analytics across multiple domains.
Network automation and orchestration are integrated into this intelligent network and utilize AI and analytics to run the network in a fully automated manner.
The aim of flexible network enablers is to increase the reliability of future networks and also create enablers that can adapt to different scenarios. The enablers investigated in Hexa-x are the following:
- Mesh ad-hoc networks provide increased coverage and capacity on a demand basis. The mesh network consists of a so-called ad hoc network controller that sets up the device-to-device (D2D) paths between devices and nodes in the area to optimize performance.
- The new 6G multi-connectivity (MC) concept aims to combine carrier aggregation and dual connectivity into one single 6G MC solution to meet the new requirements of 6G networks, including sub-THz frequencies and increased flexibility.
- NTN architecture solutions enable full global coverage.
Efficient networks enablers aim to:
- simplify the RAN and CN architecture by extending the service based architecture to the RAN by a modular, cloud compatible architecture in which the control plane functionality and common data repositories of a network are delivered by way of a set of interconnected network functions (NFs), each with authorization to access each other’s services
- reduce the signaling overhead by redesigning network functions to be more self-sustained
- increase the flexibility of the architecture through function elasticity by allowing NFs to be deployed in various cloud environments (network refactoring)
Watch the recorded Flex-top demo on Youtube.
Service management and orchestration
6G service management and orchestration (M&O) will have novel capabilities that are presented in Hexa-X deliverable D6.2 “Design of service management and orchestration functionalities” including
- Unified orchestration across the “extreme-edge, edge, core” continuum: The M&O system leverages resources from end devices to deploy 6G services. Extreme-edge nodes may have specific constraints, such as low power, limited resources shared with user-controlled applications, and volatile behavior. Thus, resource orchestration mechanisms may need to be specialized to deal with these constraints. Additionally, extreme-edge nodes often have mobility patterns that need to be considered.
- Unified management and orchestration across multiple domains: The M&O system will operate across diverse domains with various technologies and tools, owned and administered by different stakeholders, requiring converging interfaces, dynamic resource registration, and access control procedures. The system combines hierarchical and peer-to-peer federation strategies at the multi-domain coordination level.
- Increasing levels of automation: The M&O system enables functionalities such as service and network planning, design, provisioning, optimization, operation, and control with closed-loop and zero-touch solutions, reducing manual intervention. It continuously monitors network and service performance, identifying and predicting potential issues, and triggering dynamic reactions in the short or medium term.
- Adoption of data-driven and AI/ML techniques in the M&O system: The M&O system supports distributed and collaborative AI, pervasive monitoring of network performance, and scalable data and model sharing. AI/ML techniques optimize service provisioning, resource allocation, service composition, scaling, migration, reconfiguration, and optimization of resources.
- Intent-based approaches for service planning and definition. The M&O system will translate service specs and commands based on intents, including those expressed in natural language.
- Adoption of the cloud-native principles in the telco-grade environment. The M&O system for 6G networks involves using microservices for NFs, implementing a service mesh for optimal communication, and enabling deployment and updates using DevOps practices with high automation. This involves combining development and operational teams, which is a challenging but innovative approach for telco-grade environments.
Hexa-X recently published deliverable D6.3 “Final evaluation of service management and orchestration mechanisms” which presents the implementation of these novel capabilities. During EuCNC & 6G Summit 2023, Hexa-X presented demos on the “data-driven device-edge-cloud continuum management concept” to illustrate these novel concepts.
Watch the recorded demos on Youtube:
Special purpose functionalities
Hexa-X deliverable D7.2 “Special-purpose functionalities: intermediate solutions” presents technical solutions for challenging use cases with special-purpose functionalities. Three main topics are covered: flexible resource allocation in challenging environments, dependability in Industry 4.0 environments, and digital twins and novel human-machine interaction.
- Flexible resource allocation in challenging environments: This topic focuses on resource allocation in extreme environments. Such environments include remote areas and disaster zones, where traditional network infrastructure is unavailable or unreliable, or populated by mobile devices with special requirements. The objective is to design a flexible and scalable infrastructure that can quickly adapt to the changing resource demands and network conditions. The following solutions can help to reach this objective:
- In2-X communication in factory environments: Wireless data traffic in factories involves a mixture of intra-machine, inter-machine, and machine-to-infrastructure communications, requiring dynamic and intelligent spectrum sharing to minimize interference. Challenges include incompatible protocols, interference with external networks, and difficulty in aggregating real-time spectral information. Hexa-X proposes a two-level scheme for spectrum allocation: a centralized mechanism for long-term allocation and autonomous cognitive radio scheduling for short-term interference reduction.
- Radio-aware trajectory planning: Hexa-X suggests using radio-aware digital twins for intelligent trajectory planning to optimize target key performance indicators (KPIs) or key value indicators (KVIs), specifically for balancing radio performance and sustainability in industrial settings, for example, a factory floor with a private 6G network. In Hexa-X D7.2, radio-aware trajectory planning is demonstrated for optimizing unmanned aerial vehicle (UAV) flight paths, and this approach can be applied to other use cases including robots, automated guided vehicles (AGVs), and self-driving cars.
- Functional-split-aware trajectory planning for industrial vehicles: Autonomous vehicles are affected by interference from neighboring cells. Cell coordination reduces interference, but its effectiveness depends on the centralization level of the protocol stack. Dynamic adaptation of the centralization level is proposed, but connection quality at cell edges is uncertain. To address this, a mechanism is proposed for vehicles to request connection quality information to optimize trajectory planning and service provision.
- Ambient backscatter communication: recycling radio waves: Hexa-X introduces Crowd-Detectable Zero-Energy-Devices (CD-ZEDs) for utilizing radio resources in 6G networks. It uses backscattering of downlink ambient waves to broadcast the device's ID number, and a nearby smartphone connected to the network detects this message along with the ambient signals.
- Dependability in Industry 4.0 environments: This topic aims to address the requirements for dependability, reliability, and security of 6G networks in industrial environments, and includes:
- Radio resource management with a radio-aware digital twin: A radio-aware digital twin as a digital representation of the radio propagation environment is proposed, which can be used to predict the link condition between a transmitter and receiver pair.
- Dependability in UAV-assisted massive machine-type communication (MTC): The optimization of the trajectory of UAVs acting as mobile relay nodes in MTC applications aims to improve the age of information (AoI), which is a crucial metric of dependability in MTC applications relying on timely information from sensor nodes.
- Data and control plane guarantees in programmable industrial networks: 6G systems need predictable performance from both data and control planes to take advantage of virtualization and programmability. The network should forward user data and ensure timely control plane message delivery while considering the impact of virtualization and network updates.
- Network data analytics assisted by AI operations for factory network optimization: The factory network management system (FNMS) can generate and update AI/ML models that provide analytics data for joint application and RAN optimization.
- Communication-Computation-Control-Codesign (CoCoCoCo): CoCoCoCo integrates building blocks for creating dependable systems through the convergence of communication, computation, and control using a single description model instead of static approximations. It enables evaluation of error significance in terms of the impact of the errors on QoE and effective use of all available resources while adhering to additional constraints like end-to-end delay bounds.
- Digital twins and novel HMIs: Novel technologies of digital twins and human-machine interfaces (HMIs) are playing a key role in enabling new 6G use cases and establishing the convergence of the physical, digital, and human worlds. Hexa-X presented and analyzed novel HMI and DT technologies and investigated their applications in future 6G systems.
- Novel HMI for mobile human-machine and human-CPE interaction: Several types of feedback in multi-sensory HMI technologies have been named, including holographic vision, tactile, smell, and taste. Specifically, there is an intense recent interest in employing holographic vision interfaces in future 6G-driven industrial use cases, such as collaborative robots and massive twinning with humans in the loop.
- Digital-twin-empowered collaborative robots: Workplace digital twins can be enhanced in terms of localization and visualization, while a local user’s digital twin can be improved by enabling full-body pose reconstruction, using computer vision (CV) and AI techniques. This allows for more sophisticated HMI scenarios but requires additional connected hardware and may prompt a reevaluation of network requirements.
- Digital-twin-based functional split adaptation for industrial networks: The concept of a digital twin-based controller is proposed for determining the optimal functional split in a large industrial RAN that consists of heterogeneous cells. This RAN is responsible for providing connectivity to both fixed stations and moving AGVs deployed for various tasks such as transporting parts and monitoring activities within the industrial area.
- Digital twins for emergent intelligence: 6G's extensive access capacity and full coverage enable the emergence of emergent intelligence (EI) in large-scale networks. However, scaling the system and ensuring efficient communication among agents present challenges. To address this, massive twinning can be deployed, allowing the decision engine of EI agents to reside in their digital twins. Agents communicate with cloud servers hosting their digital twins, reducing signaling overhead, improving resource efficiency, and lowering latency.
At EuCNC & 6G Summit Hexa-X showed a demo on “Extreme performance in handling unexpected situations in industrial contexts” which involves virtual reality (VR), and autonomous cobots, and allows for human involvement in industrial tasks with digital twins through VR technology with immersive realistic 3D graphics. The demo showcases how to connect the human, physical, and digital worlds through cloud native resource provisioning from the cloud to the extreme edge and specifically over autonomous robots, with the human in the loop for interactions, repairs, or even manual teleoperation.
Watch the recorded demo on Youtube.
The next step – Hexa-X II
The Hexa-X project will be finalized in June 2023, and the second phase of the 6G journey which focuses on systemization and pre-standardization will continue in the Hexa-X-II project which began in January 2023.
Like what you’re reading? Please sign up for email updates on your favorite topics.Subscribe now
At the Ericsson Blog, we provide insight to make complex ideas on technology, innovation and business simple.