Source-linked AI summary
6G NeXt - Towards 6G Split Computing Network Applications: Use Cases and Architecture
Sergiy Melnyk, Qiuheng Zhou, Hans D. Schotten, Mandy Galkow-Schneider, Ingo Friese, Tobias Pfandzelter, David Bermbach, Louay Bassbouss, Alexander Zoubarev, Andy Neparidze, Arndt Kritzner, Enrico Zschau, Prasenjit Dhara, Steve Göring, William Menz, Alexander Raake, Wolfgang Rüther-Kindel, Fabian Quaeck, Nick Stuckert, Robert Vilter
TL;DR
Emerging XR applications demand data rates, low latency, and computing power beyond existing 5G capabilities, while distributed use cases require tight coordination of communications and processing. 6G NeXt proposes a joint communication-and-computing backbone with cloud-native services and resource discovery. The proposed architecture distributes processing across network levels, reducing client demands and simplifying collaborative applications.
Problem
XR data rates, transmission latency, and computation demands are expected to exceed existing 5G capabilities, while applications require communication resources to respond to computation demands and link conditions.
Source: §I · p. 1, B. AI-Assisted Communication System… · p. 5Method
The paper proposes a modular communication-and-computing backbone with cloud-native services, resource discovery and brokering, and distributed processing workflows.
Source: §III · p. 3, 2) High-performance… · p. 3, 3) Split rendering and NBMP · p. 4Results
The proposed high-performance backbone and cloud-native services distribute processing across network levels, reducing client-device demands and simplifying collaborative applications.
Source: §IV · p. 6Takeaways & Limitations
Smart Drones and HoloCom provide heterogeneous testbed scenarios for the proposed infrastructure, with the project aiming for a functional proof of concept.
Source: §IV · p. 6
Abstract
from arXiv · showhide
The definition of the sixth generation mobile communication is in the full swing. Mobile communication aims for the convergence of physical, human and digital world. Research project 6G NeXt is considering two demanding use cases, holographic communications and drones anti-collision system, which set heterogeneous requirements on the communication as well as the computing infrastructure. In both use cases, the clients are widely spread in the network and are cooperatively interacting with each other. Especially for holographic communication, also high processing power is required. This makes a high-speed distributed backbone computing infrastructure, which realises the concept of split-computing, inevitable. Furthermore, tight integration between processing facilities and wireless network is required in order to provide adequate quality of service to the users. This paper illustrates the use case scenarios and their requirements. Afterwards, an appropriate solution approach to realise those is elaborated. Here, the novel technological approaches are discussed based on the developed overall communication and computing architecture.
I. INTRODUCTION
6G NeXt addresses emerging XR demands that exceed 5G capabilities by developing a joint communication-and-computing infrastructure. It uses holographic communication and UAV anti-collision as demanding use cases to motivate distributed processing and adaptable communication.
I. INTRODUCTION
- XR data rates, transmission latency, and computation demands are expected to exceed the capabilities of existing 5G networks.The paper places 6G services and technologies in an early development phase.Source: §I · p. 1
- 6G NeXt aims to develop and evaluate a joint connectivity-and-compute infrastructure with dynamic, geo-distributed computing capabilities in mobile networks.The project extends edge computing by distributing processing facilities across backbone levels and tightly integrating them with communications.Source: §I · p. 1
- The motivating use cases are a glasses-free holographic communication platform and a real-time anti-collision system for UAVs and general aviation in shared airspace.The anti-collision system monitors aircraft positions, predicts flight paths, and detects potential collision risks.Source: §I · p. 2
- The communication stack is to adapt dynamically to application requirements and network resource distribution through agile link adaptation and automatically adjustable radio profiles.The proposed flexibility spans the application-to-physical layers and is intended to increase service efficiency, quality, and real-time capability.Source: §I · p. 2
A. UAV Collision Detection and Avoidance
Smart Drones extends a simulation-based anti-collision approach to UAVs in shared airspace. Ground-based digital twins evaluate aircraft trajectories, making timely, reliable connectivity and dynamic distribution of simulation data central requirements.
A. UAV Collision Detection and Avoidance
- Smart Drones targets drone-to-drone collision avoidance in shared airspace, with possible future extensions to manned aviation.Its goal is safe, reliable connectivity for UAVs near airports; the authors describe its approach as different from current anti-collision systems.Source: A. UAV Collision Detection and Avoidance · p. 2
- Ground-based digital twins use UAV state and flight-model data to evaluate concurrent flight paths and simulate evasive trajectories before transmitting a manoeuvre to the real UAV.The ground station gathers each UAV's altitude, position, course, velocity, and mission-profile data.Source: A. UAV Collision Detection and Avoidance · p. 2
- Offboard simulation avoids onboard computational limits, since added onboard resources increase takeoff weight and reduce mission time or range.The simulation runs at a ground-based computing station rather than on the UAV.Source: A. UAV Collision Detection and Avoidance · p. 2
- Low communication latency matters because transmission, simulation, and manoeuvre execution together must be minimized for the simulation to reflect current conditions.The passage states that more current simulation input improves the match with the real situation and system performance.Source: A. UAV Collision Detection and Avoidance · p. 2
- UAV movement across sectors requires low-latency, highly reliable connectivity and software that dynamically reallocates data and hands simulation state to other sectors.The anti-collision concept derives from the VIGA project and uses faster-than-real-time aircraft simulation to estimate trajectories and warn pilots.Source: A. UAV Collision Detection and Avoidance · p. 2
B. Holographic Communication (HoloCom)
HoloCom explores natural remote communication through true 3D holographic displays and a distributed capture-to-render workflow. Its evaluation includes both user experience and technical measures because network QoS metrics alone do not capture perceived satisfaction.
B. Holographic Communication (HoloCom)
- HoloCom aims to reconstruct a call partner as a live, true 3D hologram with natural depth and eye focus, without requiring 3D glasses.VR headsets are an optional alternative, with more limited head-and-face capture options.Source: B. Holographic Communication (HoloCom) · p. 2
- The workflow captures people in 3D on the client, synthesizes a model in the edge domain, then sends rendered views to the other client for hologram presentation.Rendered views depend on the observer's eye location, while high-quality full-3D representation also requires substantial bandwidth.Source: B. Holographic Communication (HoloCom) · pp. 2–3
- HoloCom evaluates user experience subjectively and objectively because higher QoS values do not necessarily mean higher perceived satisfaction.Measures include questionnaires, MOS, conversational and behavioral data, and objective measures of 3D rendering, encoding, and decoding.Source: B. Holographic Communication (HoloCom) · p. 3
III. 6G COMMUNICATION AND COMPUTING INFRASTRUCTURE
The paper proposes an architecture that combines communication and computing, including backbone-based distribution of computation according to application KPIs.
III. 6G COMMUNICATION AND COMPUTING INFRASTRUCTURE
- The proposed infrastructure combines communication and computing worlds in a joint architecture.The section frames the architecture alongside high-performance radio interfaces, software stacks, media processing, and AI integration.Source: §III · p. 3
A. Split Computing
- Split computing distributes computation power in the backbone according to required KPIs, extending the edge-computing concept for future applications.The passage identifies required KPIs as the basis for distributing computation.Source: A. Split Computing · p. 3
1) Cross-Layer Metrics Function:
6G NeXt proposes cross-layer metrics and a communication-and-computing backbone that use application and infrastructure requirements to guide resource discovery, placement, and task splitting.
1) Cross-Layer Metrics Function:
- Cross-layer metrics collect network, computing-layer, and application status to help applications find suitable execution, connection, or deployment environments and inform system-optimizing AI.The collected data are intended to support more efficient use of computing capacity and connectivity.Source: 1) Cross-Layer Metrics Function · p. 3
2) High-performance communication-and-computing backbone:
- The backbone allocates computing capacity and connectivity against requirements including latency, CPU/GPU availability, energy use, and cost.It comprises cloud-native extensions, the cloud layer, and the network.Source: 2) High-performance… · p. 3
- A combined Discovery and Resource Broker Function uses cross-layer measurements to match clients and services with resources, including rendering services selected for latency and price rather than physical proximity alone.The broker can locate CPU/GPU capacity together with sufficient connectivity when a service needs added capacity.Source: 2) High-performance… · p. 3
- The modular backbone replaces separate network and computing layers, enabling application-driven task splitting and decisions about where to run, connect, or deploy across the cloud continuum.The approach complements the 3GPP 5G edge-cloud specification.Source: 2) High-performance… · p. 3
3) Split rendering and NBMP:
For split rendering, 6G NeXt uses NBMP to describe media-processing workflows and distribute their functions across available infrastructure according to processing and location requirements.
3) Split rendering and NBMP:
- NBMP supports split rendering by enabling media-processing workflows to be defined and distributed across the available infrastructure.The standard provides schemes for describing functions and workflows and interfaces among the involved components.Source: 3) Split rendering and NBMP · pp. 3–4
- A central controller can use workflow processing or location requirements and resource-broker information to deploy media-processing functions across the infrastructure.The paper identifies on-demand distributed deployment of a predefined processing pipeline as a benefit for holographic communication.Source: 3) Split rendering and NBMP · p. 4
- In the illustrated pipeline, preprocessing stays near capture, the next three functions go to the nearest edge, and decoding is deployed near the communication partner.The standardized description allows the workflow to be parameterized, reused, extended, or modified.Source: 3) Split rendering and NBMP · p. 4
4) Edge-to-Cloud Serverless Platform:
The edge-to-cloud platform combines serverless compute, geo-distributed data management, and decentralized messaging to support adaptive application services across distributed resources.
4) Edge-to-Cloud Serverless Platform:
- The platform adopts serverless cloud concepts for computation, data management, and communication across the edge-cloud continuum, with adaptive and scalable abstractions integrating geo-distributed edge resources and cloud bursting.The high-speed backbone and embedded compute resources are paired with these software abstractions.Source: 4) Edge-to-Cloud Serverless Platform · p. 4
- FaaS uses small, event-driven, stateless functions for fine-grained resource allocation and dynamic request routing, while edge deployments must limit abstraction overhead.The paper notes that lightweight edge FaaS platforms such as tinyFaaS can run on constrained edge devices.Source: 4) Edge-to-Cloud Serverless Platform · p. 4
- Because target applications such as UAV collision avoidance need persistent trajectory and capability data, function migration is supported by automated data migration across geo-distributed edge-to-cloud resources.Application-controlled replicas are combined with platform-side placement and migration to reduce access latency, distribute data consistently, and adhere to privacy requirements.Source: 4) Edge-to-Cloud Serverless Platform · pp. 4–5
- A decentralized, geo-distributed Pub/Sub message bus provides edge-to-edge communication, including between heterogeneous UAVs, without routing all messages through the cloud.GeoBroker is cited as a prototype that adds geographical context so messages reach local subscribers.Source: 4) Edge-to-Cloud Serverless Platform · p. 5
B. AI-Assisted Communication System Approach
The communication approach uses predictive analytics to adapt radio resources to application demands and changing link conditions, while recognizing that wireless prediction remains uncertain.
B. AI-Assisted Communication System Approach
- Radio resource management must respond to application computation demands and environmental changes in link quality to integrate communication with split computing.The research focus is predictive algorithms that provide adjustable and guaranteed QoS parameters for participants.Source: B. AI-Assisted Communication System… · p. 5
1) AI-based system optimisation:
- ML models can predict QoS and radio KPIs, including radio environment maps, channel distribution information maps, and spectral efficiency, to support proactive application and radio-access-network adaptation.Predictions are intended to account in advance for spatiotemporal wireless dynamics.Source: 1) AI-based system optimisation · p. 5
- Predicted QoS can help resource-management schemes offer more reliable future guarantees, although those guarantees cannot always be met; channel-prediction methods are tested using real communication data and software-defined-radio measurements.Prediction uncertainty is unavoidable under stochastic, non-stationary signal propagation, and can be increased by inefficient training data, uncertain inputs, and limited model generalizability.Source: 1) AI-based system optimisation · p. 5
- The implemented communication platform relies on ML methods to improve system throughput and prioritize users expected to experience poor performance.Source: 1) AI-based system optimisation · p. 5
2) Agile Coding Schemes:
The section considers agile coding and link distribution to meet heterogeneous traffic requirements, adapting coding rates and distributing data across communication links.
2) Agile Coding Schemes:
- Heterogeneous application traffic creates distinct link QoS requirements, and some data types may benefit from using several links.The network should provide agility to tailor communication to each traffic type.Source: 2) Agile Coding Schemes · p. 5
- Fountain codes are rateless K-out-of-N schemes whose coding rate can be adjusted by generating an unrestricted number of sub-codewords.They require an erasure channel: they tolerate packet loss but cannot recognize corrupted packets; combining error-detecting codes with fountain codes gains performance.Source: 2) Agile Coding Schemes · p. 5
- AI-based channel prediction can adapt coding rates to changing radio conditions to meet packet-error requirements and serve different traffic classes.Source: 2) Agile Coding Schemes · p. 5
- Fountain-code sub-codewords can be distributed across links based on link state; using different radio access technologies improves resilience and can support seamless handover.Traffic can be adaptively distributed across links to improve coverage for a single radio access technology.Source: 2) Agile Coding Schemes · p. 5
3) Testbed System:
The testbed places processing across cloud, edge, and user equipment, with portable setups for laboratory and open-field use and a flight-capable communication module for Smart Drones.
3) Testbed System:
- The project attaches distributed processing intelligence to cloud, edge, and user-equipment layers, supported by transportable laboratory and open-field test stands.Source: 3) Testbed System · pp. 5–6
- Test stands use the OpenAirInterface software stack, mixed SRD gNB and UE hardware from the USRP series, and commercial modems in the 3.7–3.8 GHz band.Source: 3) Testbed System · p. 6
- The setup is intended to test use-case latency and bandwidth realistically with fewer subscribers than commercial installations.Source: 3) Testbed System · p. 6
- Smart Drones flight testing requires a lightweight, low-power terminal with multiconnective communication, sensor connections, and wired flight-controller communication.A module derived from NetMobilBox provides this functionality, building on development from the 5G NetMobil project.Source: 3) Testbed System · p. 6
IV. CONCLUSION
The paper presents a joint communication-and-computing architecture and uses Smart Drones and HoloCom as heterogeneous testbeds for pursuing a functional proof of concept.
IV. CONCLUSION
- A high-performance backbone and cloud-native services distribute processing across network levels, reducing client-device demands and simplifying collaborative applications.Source: §IV · p. 6
- Smart Drones and HoloCom provide testbed scenarios with heterogeneous requirements for latency, throughput, reliability, and computation availability.Source: §IV · p. 6
- The project aims to move beyond conceptual evaluation toward a functional proof of concept based on the selected use cases.Source: §IV · p. 6