Maximising Energy Efficiency in Large-Scale Open RAN: Hybrid xApps and Digital Twin Integration
Ahmed Al-Tahmeesschi 1,* , Yi Chu 1,* , Gurdeep Singh 2 , Charles Turyagyenda 2 , Dritan Kaleshi 2 , David Grace 1 , Hamed Ahmadi 1 1 School of Physics, Engineering and Technology, University of York, United Kingdom 2 Digital Catapult, United Kingdom · 2025
Read Maximising Energy Efficiency in Large-Scale Open RAN: Hybrid xApps and Digital Twin Integration by Ahmed Al-Tahmeesschi 1,* , Yi Chu 1,* , Gurdeep Singh 2 , Charles Turyagyenda 2 , Dritan Kaleshi 2 , David Grace 1 , Hamed Ahmadi 1 1 School of Physics, Engineering and Technology, University of York, United Kingdom 2 Digital Catapult, United Kingdom (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-10097.txt.
Introduction
Ahmed Al-Tahmeesschi and 6 co-authors tackle the power bill of 5G networks in this 2025 paper. Radio access networks use much of a mobile system's energy, and Open RAN designs, which split the network into parts from different vendors, make power harder to manage. The team built a hybrid control app that mixes simple rules with unsupervised learning to switch radio units into sleep modes.
A digital twin of the network, fed by a scenario generator, let them test the app on a large, realistic setup before touching live equipment. The controller saved about 13 percent of energy while keeping user quality intact. The authors argue the mix of heuristics, learning, and a twin makes the savings safe to deploy.
Catalog record
| CREATOR | Ahmed Al-Tahmeesschi 1,* , Yi Chu 1,* , Gurdeep Singh 2 , Charles Turyagyenda 2 , Dritan Kaleshi 2 , David Grace 1 , Hamed Ahmadi 1 1 School of Physics, Engineering and Technology, University of York, United Kingdom 2 Digital Catapult, United Kingdom |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 32 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-10097
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Maximising Energy Efficiency in Large-Scale Open RAN: Hybrid xApps and Digital Twin Integration Ahmed Al-Tahmeesschi 1,* , Yi Chu 1,* , Gurdeep Singh 2 , Charles Turyagyenda 2 , Dritan Kaleshi 2 , David Grace 1 , Hamed Ahmadi 1 1 School of Physics, Engineering and Technology, University of York, United Kingdom 2 Digital Catapult, United Kingdom Abstract The growing demand for high-speed, ultra-reliable, and low-latency communications in 5G and beyond networks has significantly driven up power consumption, particularly within the Radio Access Network (RAN). This surge in energy demand poses critical operational and sustainability challenges for mobile network operators, necessitating innovative solutions that enhance energy efficiency without compromising Quality of Service (QoS). Open Radio Access Network (O-RAN), spearheaded by the O-RAN Alliance, offers disaggregated, programmable, and intelligent architectures, promoting flexibility, interoperability, and cost-effectiveness. However, this disaggregated approach adds complexity, particularly in managing power consumption across diverse network components such as Open Radio Units (RUs). In this paper, we propose a hybrid xApp leveraging heuristic methods and unsupervised machine learning, integrated with digital twin technology through the TeraVM AI RAN Scenario Generator (AI-RSG). This approach dynamically manages RU sleep modes to effectively reduce energy consumption. Our experimental evaluation in a realistic, large-scale emulated Open RAN scenario demonstrates that the hybrid xApp achieves approximately 13 % \% energy savings, highlighting its practicality and significant potential for real-world deployments without compromising user QoS. * * footnotetext: These authors contributed equally to this work. I Introduction The rapid evolution of wireless communication technologies necessitates innovative approaches to meet the increasing demands for throughput, coverage, and user experiences. However, there are critical environmental impacts arise together with the increasing demands such as energy consumption and carbon footprint. Throughout the generations of the mobile networks, the RAN ( RAN ) has always been an imperative component and the direct gateway of connecting the mobile UE ( UE ) over the air. The RAN includes the computing infrastructure hosting the heavy baseband signal processing as well as power-hungry hardware components (such as the power amplifiers), therefore, its energy consumption has always been a major concern. From a decade ago the MNO have already been reported as one of the top energy consumers [ 1 ] . The concern has not been resolved with the deployment of the 5G ( 5G ) [ 2 ] . To address the increasing demands for network capacity, coverage and latency, mass deployment of ultra-dense small-scale BS , known as network densification, is the major trend of 5G and future networks. Such HDD ( HDD ) inevitability increases mobile network energy consumption, leading to a greater carbon footprint and higher operating cost for the MNO . The most well-established architecture of Open RAN is O-RAN ( O-RAN ) which was proposed by the O-RAN Alliance [ 3 ] . The benefits of O-RAN include standardized open interfaces that allow multi-vendor network deployment [ 3 ] and the integrated AI ( AI )/ ML ( ML ) hosted in the RIC ( RIC ) [ 4 ] . O-RAN disaggregates the RAN into a RU ( RU ), a DU ( DU ) and a CU ( CU ) with each unit hosting a certain set of RAN functions according to different functional split options [ 5 ] . The well-defined interfaces allow interoperability across RU , DU and CU from multiple vendors which lowers the difficulty for small and medium manufacturers to enter the RAN market, therefore, fostering the market competitiveness, innovation and upgrade cycles [ 6 ] . The RIC on the other hand, acts as the central intelligence of the RAN which conducts various AI / ML based network performance optimizations via specific interfaces such as E2 [ 7 ] and O1 [ 8 ] . The widely supported E2 interface specifies the messages [ 9 ] between the RAN and the Near-RT ( Near-RT ) RIC which handles operations requiring a latency of between 0.1 and 1 second [ 7 ] . The customisable xApps hosted at the Near-RT RIC allow the MNO to monitor live network performance metrics via the E2SM-KPM ( E2SM-KPM ) [ 10 ] and to change the network parameters and configurations via the E2SM-RC ( E2SM-RC ) [ 11 ] and E2SM-CCC ( E2SM-CCC ) [ 12 ] to optimize the network performance and UE QoS ( QoS ). Such standardized methods for monitoring and controlling the network have attracted many interests from the academia and industry and resulted in several optimization directions such as energy efficiency [ 13 ] , traffic steering [ 14 ] and network slicing [ 15 ] . In this paper, we address the energy efficiency optimization assisted by a powerful industrial RAN emulation and RIC testing tool, the TeraVM AI RSG ( RSG ) provided by VIAVI [ 16 ] . This tool creates a DT ( DT ) of the RAN that simulates system level behaviour of the network with scalable and flexible deployment options such as a large number of UE and cells, 3GPP standardized propagation models, UE mobility and traffic profiles, cell RF ( RF ) and energy models, MAC ( MAC ) scheduler algorithms, etc. The RAN nodes can be controlled during live simulations via the REST API ( API ) or E2 messages with actuations such as switching cells on and off and issuing HO ( HO ) commands. The RSG also offers exposing the network KPM ( KPM ) reports and RC ( RC ) commands to external IPs via the E2 interface which makes it the ideal tool for testing RIC and xApp development. Later in this paper we will explain in detail how we configure this tool to generate a DT for a large-scale network and the amount of energy saving achieved with our AI / ML powered xApp. The contributions of this work are as follows: • We propose a novel hybrid Energy Saving xApp that integrates heuristic rules and unsupervised machine learning for intelligent O-RU sleep control within Open RAN. • We leverage DT via the VIAVI TeraVM AI-RSG to emulate large-scale Open RAN networks with realistic user mobility and channel propagation conditions. • We design lightweight clustering-based mechanisms to identify and activate suitable sleeping cells to meet dynamic user demands, while switching off underutilised RUs with minimal computational overhead. • We validate our approach in a dense urban-like environment with 246 UEs and 51 cells, demonstrating up to 13% energy savings without compromising user QoS. II Architectural Design Fig. 1 shows the high-level architectural design of the system, the interactions between the proposed ES ( ES )-xApp and the AI RSG (RIC Tester) as well as the submodules of the xApp. The functionalities of the submodules are described below. Fig. 1 : High-level architecture and submodules • REST API : this submodule manages interactions between the xApp and the RIC tester via HTTP requests sent to specific endpoints and corresponding responses. Its detailed functionalities are described in submodules below. • Simulation Automation: this submodule has two interactions with the RIC tester including starting the network simulation using HTTP POST (with a payload) with a specific network configuration (json format) and stopping the network simulation using HTTP