BeGREEN will take a holistic view to propose evolving radio networks that not only accommodate increasing traffic and services but also consider power consumption as a factor.
BeGREEN will perform research and implementation at different levels, i.e., at the hardware level, at the link level, and at the system level toward improved energy efficiency at radio access design. The AI/ML techniques are used to reduce the required processing and also to recognise patterns in the system level data associated with the behaviour of the user base and to learn the most appropriate response to this behaviour in terms of both network performance and also energy consumption. BeGREEN uses the emerging O-RAN standard and its evolution as the baseline architecture. The disaggregation, virtualisation and network and service management capabilities inherent in O-RAN provide the mechanisms to realise many of the infrastructure changes and techniques for energy optimisation discussed above.
A priority for BeGREEN will be to establish a solid techno-economic basis for assessment of technology choices where energy efficiency is an explicit characteristic rather than an afterthought. With 5G networks being rolled out across the globe, and technologies addressing 6G use cases being actively debated in academic and industrial fora, there is still surprisingly little analysis of the actual impact of specific engineering choices made by the technical community in development of 5G. BeGREEN takes an evidence-based assessment of current and emerging radio access technology choices to expose the practical energy cost of cutting-edge technologies to help community achieve energy consumption targets. With this fundamental capability in place, the project will be in a strong position to achieve the objectives listed below.
Vision
Concept
BeGREEN evaluate different mechanisms by which power consumption could be reduced based on the following pillars:
At the architecture level, planning and evaluation of a massive MIMO included RAN design to achieve flexible and energy efficient connectivity considering spectrum utilisation, interference mitigation and architecture/processing complexity.
At the hardware and infrastructure level, radio-unit controlling schemes will be used in power amplifiers energy optimisation. Also, an offloading engine for hardware acceleration will be employed to achieve energy efficiency when performing radio access functions and network function virtualisation.
At the link level, the integrated sensing techniques are used to provide a better estimate of the impact of the radio channel toward improvements in spectral efficiency against the increased power consumption associated with the resulting calculations.
At the system level, the project pursues the development and evaluation of AI-based procedures to adapt the energy consumption of softwarised network functions, aiming to minimize the overall consumed energy according to the utilisation patterns of network. BeGREEN proposes an "Intelligent Plane", as an additional plane along with user plane and data plane, that allows the data, model and inference to be seamlessly exchanged between network functions.
BeGREEN innovations targeting RAN energy efficiency enhancement include:
o Introducing a GPU-based offloading engine for acceleration of DU and RU GPU based for mMIMO radio-access heavy data processing
o O-RAN based RU energy consumption optimization including, i) RU controlling schemes from O-RAN rApps, and xApps, ii) AI-based DPD and envelope tracking solutions
o Introducing new methods in RIS aided sensing for assisted radio access
o Introducing an ORAN -based 'Intelligent Plane' for AI-assisted network function energy optimisation
o Introducing new ML algorithms to dynamically select CPU power saving modes (e.g., C-states) or adapt the number of active VNF instances to minimize energy consumption without affecting workload performance
o Introducing eXplainable and interpretable AI algorithms to accurately pinpoint energy influencing factors of the network functions beyond traffic
Innovation
Demonstrators
BeGREEN technologies will be showcased in three demonstrations:
At IHP premises, the joint communications and sensing techniques toward efficient resource allocation and optimised power consumption will be showcased. The project targets using reconfigurable intelligent surfaces for energy saving scenarios in the demonstrations.
Furthermore, two major project demonstrations will be performed at BT premises in Adastral park:
First, the 'Intelligent Plane' implemented using ORAN rApps and xApps will be demonstrated on a network emulator.
Then, the project final integrated demonstrator using the Adastral testbed facilities to showcase BeGREEN technology innovations.
BeGREEN pursues the following objectives:
Planning and evaluation of a reference mMIMO architecture to achieve:
<> Energy efficient connectivity;
<> Proper spectrum resource allocation;
<> Optimised resource allocation;
<> Balance of energy transmission and interference mitigation;
'DU/RU offloading engine' GPU based acceleration as a fundamental technology component to achieve energy efficiency for high computing offloading platform that enables mMIMO radio-access heavy data processing, CU-UP hardware acceleration of PDCP, and acceleration of the RIC xApps;
O-RAN based RU energy consumption optimization including, i) RU controlling schemes from O-RAN rApps, and xApps, ii) AI-based DPD and envelope tracking solutions;
Development of a sensing-assisted communication approach, enabling optimal spatial resource-allocation, e.g., enhanced transmit power and beamwidth allocation, towards network energy usage optimization;
Design and development of an 'Intelligent Plane', incorporating O-RAN, along with user plane and data plane, for AI-assisted network function energy optimisation that allows the data, model and inference to seamlessly exchange over the network;
Design AI/ML algorithms that dynamically select CPU power saving modes (e.g., C-states) or adapt the number of active VNF instances to minimize energy consumption without affecting workload performance;
Design AI/ML algorithms that uses eXplainable and interpretable AI algorithms (e.g.: Shapley algorithms and partial dependency plots) that will accurately pinpoint energy influencing factors of the network functions beyond traffic;
Design and implementation of next generation Edge, aiming to minimize the overall energy cost by using, i) AI-assisted procedures to jointly control RAN resources and Edge service parameters, ii) RAN user-plane NF acceleration plus EAS offload.
