Robust and AI Native 6G for Green Networks
Project Status: set-up
Start Date: January 2024
End Date: December 2026
Budget (total): 9829.76 K€
Effort: 110.71 PY
Name: Cicek Cavdar
Company: Royal Institute of Technology
CityPassenger SA, France
Institut Mines Télécom, France
Instituto Superior de Engenharia de Lisboa, Portugal
Instituto Superior de Engenharia do Porto (ISEP/CISTER), Portugal
Royal Institute of Technology, KTH (Kungliga Tekniska Högskolan), Sweden
Tele2 Sverige AB, Sweden
Infovista AB, Sweden
MIC Nordic AB, Sweden
Maven Wireless, Sweden
PI Works, Turkey
The roll-out of 5G is ramping up in many countries and although 5G technology has a much more energy-efficient data transfer than legacy networks, it consumes much more energy due to increased data volumes. The rise of industrial and sensitive applications in 5G networks increases at the same time the resilience requirements, and there is a trade-off between resiliency and energy-efficiency since extra resources are needed to guarantee resiliency of the network against changing channel conditions, link failures and disasters. This trade-off is based on the legacy cellular network architecture not allowing much elasticity based on fixed network coverage and deployment. Besides this tradeoff, the current cloud-centric data processing and storage architecture is an obstacle against both resiliency and energy-efficiency, as the need for sending data from devices to a far data centre results in a vulnerability against any intermediate link failure, and incurs high energy costs on the full path.
Recent advancements in virtualisation, softwarization and cloudification of network resources enables us to design cloud-based cell-less networks where processing and network resources can be dynamically reconfigured following the changing traffic conditions. Both resilient and energy-efficient networks by design are then possible if advanced network technologies at the physical layer can be adaptively controlled by a fully elastic, cloud and AI native network architecture. These technologies also allow the deployment of small data centres at the network edge, for storing and processing local data, and the orchestration of the processing, storage and networking resources on the continuum between devices and the central cloud.
The rise of renewable energy resources, distributed over wide areas and co-located with mobile network infrastructures and edge data centres offers a great opportunity for exploiting the surplus of energy production in local data centers, for processing local data, in a federated learning approach.
Hence, the key issues to be addressed by Robust and AI Native 6G for Green Networks (RAI-6Green) are
- Proposal of AI-based Network Energy Efficiency and Spectral Efficiency assessment and optimization including the access, the point-of-presence and the core networks.
- Explore solutions for accelerating computation facilities that empower AI in a distributed or centralized architecture. This item is closely related to green computing facilities.
- Solutions for control procedures that optimize networks sites power consumption and power usage (including local solar energy) depending on a variety of parameters that could be non-correlated
- Proposal of risk-sensitive (delay, robustness, energy consumption, quality of service…) optimization for performance management that takes into account the end-to-end path.
- Proposal of smart, autonomous and parameter-free BSs that learn from the environment and activate the relevant energy saving features when needed with the optimized parameter settings.
- Definition of AI-based Network multi-objective optimization using data (traffic prediction, resource preallocation, self-healing).
- Definition of KPIs for robustness and energy efficiency of network slices and adequate measurement and reporting methods (for energy efficiency standards evolutions).
- Interconnection of 6G networks, their edge data centers and the smart grid for an opportunistic exploitation of the surplus of energy production within the edge data centers, for AI as a Service data processing.
- Technical-economic study for AI-based, green deployment of Edge computing, and proposal for coinvestment plan involving stakeholders, such as network operator and service providers.
The project will also consider services based on heterogeneous data collection. The main purpose is to study new business services not only for enhancing the connectivity but also to build business models based on AI such as energy as a service (EaaS), connectivity marketplaces, and smart cities. Indeed, new application domains and territories are counting on telecom networks to facilitate their usage. The 5th generation of telecom networks is built to address those verticals and they will generate an amount of variable and uncorrelated data that we aim to use as inputs to orchestrate their behaviour.
The main goal of this project is to achieve an improvement of about 30-40% of the end-to-end energy efficiency compared to current networks. This target is very challenging but possible considering results obtained by IT industries for operating data centers (in cooling for example) and access networks (e.g. resources optimization) for instance. These network segments are the most energy consuming parts of the IT infrastructure.