Periodic Reporting for period 1 - ENGCoN (Design and development of energy-efficient next-generation communication networks)
Período documentado: 2024-06-01 hasta 2026-05-31
Resumen del contexto y de los objetivos generales del proyecto
The ENGCoN project, “Design and Development of Energy-Efficient Next-Generation Communication Networks”, addresses a central challenge facing future digital infrastructure: the rapid growth in connectivity, data traffic, artificial intelligence-enabled services, and 5G/6G network capabilities is increasing the energy burden of communication systems. Future networks must deliver higher bandwidth, lower latency, greater reliability, and flexible service support, but these advances risk intensifying the “energy crunch” unless energy efficiency is treated as a core design requirement rather than a secondary optimisation target.
The motivation for the project is technical as well as societal. Technically, next-generation networks require new models, benchmarks, and design methods that can capture energy use across radio access, optical transport, edge/cloud computing, virtualised Open RAN, and AI-enabled network control. Existing approaches often treat these domains separately or rely on incomplete assumptions about equipment trends, workload behaviour, and emerging architectures. Societally, communication networks are part of the critical infrastructure required for digital services, but their future deployment must remain compatible with climate, sustainability, and resilience objectives. In this sense, ENGCoN is aligned with the European policy context supporting green, secure, and resilient digital infrastructure, and with the MSCA Green Charter’s emphasis on embedding sustainability into research practice.
The overall objective of ENGCoN is to develop knowledge, models, and design principles that support more energy-efficient next-generation communication networks. The project combines top-down and bottom-up energy analyses to obtain a more complete view of energy use and trends. It considers equipment evolution and semiconductor technology trends, develops models for emerging technologies such as 5G/6G core networks, virtualised Open RAN, edge/cloud systems, federated artificial intelligence, and optical communication systems, and uses experimental and research-infrastructure contexts such as OpenIreland and COSMOS to strengthen the relevance of the work. The project also aims to identify benchmarks, trade-offs, and sustainability challenges that can guide future research, network design, and deployment strategies.
The pathway to impact is primarily knowledge-based. ENGCoN is expected to contribute through peer-reviewed publications, conference papers, invited talks, engagement with research infrastructure, open-science routes where possible, and continued collaboration with academic, testbed, and industry-facing stakeholders. Its outputs are intended to be reusable by researchers working on 5G/6G, Open RAN, optical networks, wireless systems, edge/cloud platforms, AI for networks, and sustainable network operation. For the telecom and ICT sector, the project provides evidence relevant to energy-aware deployment, power-consumption modelling, virtualised architectures, and sustainable AI-enabled services. For policy and sustainability stakeholders, it helps clarify the energy implications of future digital infrastructure and supports a more transparent assessment of green-network strategies.
The expected impact is significant at the scale of an MSCA Postdoctoral Fellowship, while remaining realistic. ENGCoN does not claim immediate deployment-level industrial impact by itself. Instead, its expected contribution is to strengthen the evidence base needed for future demonstrations, comparative benchmarks, network-energy assessment methods, and follow-up research proposals on green digital infrastructure. The project’s impact pathway also includes education and public engagement: accessible summaries, outreach activities, seminars, and student-facing engagement help explain why the energy efficiency of communication networks matters for society.
ENGCoN sets out to address a critical need that future communication networks must support growing digital demand without unsustainable increases in energy consumption. By developing energy models, benchmarking methods, experimental insights, and dissemination routes for expert and non-specialist audiences, the project contributes to the broader transition towards sustainable, intelligent, and resilient communication infrastructure.
The motivation for the project is technical as well as societal. Technically, next-generation networks require new models, benchmarks, and design methods that can capture energy use across radio access, optical transport, edge/cloud computing, virtualised Open RAN, and AI-enabled network control. Existing approaches often treat these domains separately or rely on incomplete assumptions about equipment trends, workload behaviour, and emerging architectures. Societally, communication networks are part of the critical infrastructure required for digital services, but their future deployment must remain compatible with climate, sustainability, and resilience objectives. In this sense, ENGCoN is aligned with the European policy context supporting green, secure, and resilient digital infrastructure, and with the MSCA Green Charter’s emphasis on embedding sustainability into research practice.
The overall objective of ENGCoN is to develop knowledge, models, and design principles that support more energy-efficient next-generation communication networks. The project combines top-down and bottom-up energy analyses to obtain a more complete view of energy use and trends. It considers equipment evolution and semiconductor technology trends, develops models for emerging technologies such as 5G/6G core networks, virtualised Open RAN, edge/cloud systems, federated artificial intelligence, and optical communication systems, and uses experimental and research-infrastructure contexts such as OpenIreland and COSMOS to strengthen the relevance of the work. The project also aims to identify benchmarks, trade-offs, and sustainability challenges that can guide future research, network design, and deployment strategies.
The pathway to impact is primarily knowledge-based. ENGCoN is expected to contribute through peer-reviewed publications, conference papers, invited talks, engagement with research infrastructure, open-science routes where possible, and continued collaboration with academic, testbed, and industry-facing stakeholders. Its outputs are intended to be reusable by researchers working on 5G/6G, Open RAN, optical networks, wireless systems, edge/cloud platforms, AI for networks, and sustainable network operation. For the telecom and ICT sector, the project provides evidence relevant to energy-aware deployment, power-consumption modelling, virtualised architectures, and sustainable AI-enabled services. For policy and sustainability stakeholders, it helps clarify the energy implications of future digital infrastructure and supports a more transparent assessment of green-network strategies.
The expected impact is significant at the scale of an MSCA Postdoctoral Fellowship, while remaining realistic. ENGCoN does not claim immediate deployment-level industrial impact by itself. Instead, its expected contribution is to strengthen the evidence base needed for future demonstrations, comparative benchmarks, network-energy assessment methods, and follow-up research proposals on green digital infrastructure. The project’s impact pathway also includes education and public engagement: accessible summaries, outreach activities, seminars, and student-facing engagement help explain why the energy efficiency of communication networks matters for society.
ENGCoN sets out to address a critical need that future communication networks must support growing digital demand without unsustainable increases in energy consumption. By developing energy models, benchmarking methods, experimental insights, and dissemination routes for expert and non-specialist audiences, the project contributes to the broader transition towards sustainable, intelligent, and resilient communication infrastructure.
Trabajo realizado desde el comienzo del proyecto hasta el final del período abarcado por el informe y los principales resultados hasta la fecha
The technical and scientific work of ENGCoN focused on developing methods, models, and experimental insight for improving the energy efficiency of next-generation communication networks. The work concentrated on three connected technical directions: power-consumption modelling for virtualised and Open RAN systems, energy-aware design of multi-layer and multi-domain communication-network architectures, and machine-learning-based digital-twin modelling for optical communication systems.
The first major activity was the development and analysis of power-consumption models for Open RAN and virtualised radio-access-network architectures. This work addressed the need to understand how disaggregated, software-based, and cloudified radio access networks consume energy under different architectural and operational conditions. The project developed modelling approaches for estimating and predicting power consumption in virtualised O-RAN systems, including machine-learning-based predictive models. The outcome was a stronger quantitative basis for comparing O-RAN design choices, identifying energy-relevant parameters, and supporting future work on energy-aware deployment and operation of virtualised RAN infrastructure. This work resulted in published conference papers on power-consumption models for Open-RAN architectures and machine-learning-based predictive models for power consumption in virtualised O-RANs.
The second technical activity was the study of energy-efficient and carbon-aware network architecture design. The project investigated architectural approaches for multi-layer computing-power networks, network slicing, topology abstraction, and multi-domain routing. This work examined how communication, computing, and routing decisions can be represented in a form suitable for energy- and carbon-aware optimisation. The main outcome was the development of modelling and abstraction methods that can support energy-aware resource allocation and routing decisions in future network architectures. This activity produced a published journal paper on self-adaptive auxiliary cube methods for multi-tenant slicing in multi-layer computing power networks, and further work on integrated self-adaptive topology abstraction for carbon-efficient multi-domain computing-aware routing was progressed.
The third scientific activity addressed machine-learning-based modelling of optical communication systems for digital-twin applications. The project developed and applied learning-based methods for predicting Raman tilt and stimulated Raman scattering behaviour in dense wavelength-division multiplexed and ROADM-based optical transmission systems. These models support the creation of more accurate digital twins of optical links and transmission systems. The outcome was a set of predictive modelling methods that can help estimate physical-layer behaviour without relying only on repeated direct measurements or purely analytical approximations. This work resulted in published conference papers on Raman tilt prediction for digital-twin modelling of ROADM-based transmission systems and CNN-based transfer learning for stimulated Raman scattering spectrum and tilt prediction.
The project also carried out testbed-oriented technical work linked to OpenIreland and COSMOS. The activity used the experimental and research infrastructure available through TCD/CONNECT and the international collaboration with Columbia University to connect the modelling work to realistic network environments. The outcome was practical experience applying energy measurement and modelling concepts in programmable and experimental network settings, including optical, wireless, and edge/cloud scenarios. This strengthened the relevance of the modelling work and provided a basis for follow-up testbed-based energy-efficiency studies.
The main scientific achievements of the project were: one published journal paper, four published conference papers, one journal paper under review, and a further set of drafted manuscripts on O-RAN power modelling, sleep-mode-aware energy modelling, energy-latency trade-offs with distributed baseband processing and AI inference, energy savings through resource sharing, and cross-platform power-consumption prediction using transfer learning. These outputs show that the project moved from the initial goal of understanding energy trends in future networks towards concrete models, predictive methods, and architectural analyses for Open RAN, optical systems, and multi-domain energy-aware network design.
The main outcomes of the actions are reusable scientific knowledge, validated modelling approaches, and a clearer technical basis for benchmarking energy use in next-generation communication networks. The work contributes to future research on 5G/6G, Open RAN, optical networks, edge/cloud architectures, and AI-enabled network management by providing models and methods that other researchers can compare, extend, and apply in future experimental and simulation studies.
The first major activity was the development and analysis of power-consumption models for Open RAN and virtualised radio-access-network architectures. This work addressed the need to understand how disaggregated, software-based, and cloudified radio access networks consume energy under different architectural and operational conditions. The project developed modelling approaches for estimating and predicting power consumption in virtualised O-RAN systems, including machine-learning-based predictive models. The outcome was a stronger quantitative basis for comparing O-RAN design choices, identifying energy-relevant parameters, and supporting future work on energy-aware deployment and operation of virtualised RAN infrastructure. This work resulted in published conference papers on power-consumption models for Open-RAN architectures and machine-learning-based predictive models for power consumption in virtualised O-RANs.
The second technical activity was the study of energy-efficient and carbon-aware network architecture design. The project investigated architectural approaches for multi-layer computing-power networks, network slicing, topology abstraction, and multi-domain routing. This work examined how communication, computing, and routing decisions can be represented in a form suitable for energy- and carbon-aware optimisation. The main outcome was the development of modelling and abstraction methods that can support energy-aware resource allocation and routing decisions in future network architectures. This activity produced a published journal paper on self-adaptive auxiliary cube methods for multi-tenant slicing in multi-layer computing power networks, and further work on integrated self-adaptive topology abstraction for carbon-efficient multi-domain computing-aware routing was progressed.
The third scientific activity addressed machine-learning-based modelling of optical communication systems for digital-twin applications. The project developed and applied learning-based methods for predicting Raman tilt and stimulated Raman scattering behaviour in dense wavelength-division multiplexed and ROADM-based optical transmission systems. These models support the creation of more accurate digital twins of optical links and transmission systems. The outcome was a set of predictive modelling methods that can help estimate physical-layer behaviour without relying only on repeated direct measurements or purely analytical approximations. This work resulted in published conference papers on Raman tilt prediction for digital-twin modelling of ROADM-based transmission systems and CNN-based transfer learning for stimulated Raman scattering spectrum and tilt prediction.
The project also carried out testbed-oriented technical work linked to OpenIreland and COSMOS. The activity used the experimental and research infrastructure available through TCD/CONNECT and the international collaboration with Columbia University to connect the modelling work to realistic network environments. The outcome was practical experience applying energy measurement and modelling concepts in programmable and experimental network settings, including optical, wireless, and edge/cloud scenarios. This strengthened the relevance of the modelling work and provided a basis for follow-up testbed-based energy-efficiency studies.
The main scientific achievements of the project were: one published journal paper, four published conference papers, one journal paper under review, and a further set of drafted manuscripts on O-RAN power modelling, sleep-mode-aware energy modelling, energy-latency trade-offs with distributed baseband processing and AI inference, energy savings through resource sharing, and cross-platform power-consumption prediction using transfer learning. These outputs show that the project moved from the initial goal of understanding energy trends in future networks towards concrete models, predictive methods, and architectural analyses for Open RAN, optical systems, and multi-domain energy-aware network design.
The main outcomes of the actions are reusable scientific knowledge, validated modelling approaches, and a clearer technical basis for benchmarking energy use in next-generation communication networks. The work contributes to future research on 5G/6G, Open RAN, optical networks, edge/cloud architectures, and AI-enabled network management by providing models and methods that other researchers can compare, extend, and apply in future experimental and simulation studies.
Avances que van más allá del estado de la técnica e impacto potencial esperado (incluida la repercusión socioeconómica y las implicaciones sociales más amplias del proyecto hasta la fecha)
ENGCoN produced scientific results in three main areas: energy modelling and benchmarking for future communication networks; power-consumption modelling for virtualised Open RAN and related architectures; and machine-learning-based digital-twin modelling for optical communication systems. These results address the need for more accurate, reusable, and experimentally grounded methods for assessing the energy implications of 5G/6G, Open RAN, edge/cloud, AI-enabled, and optical network systems.
The first group of results concerns energy models and benchmarking methods for future networks. The project developed methods and architectural analyses that support the comparison of energy use across network components, computing resources, and multi-domain communication systems. A published journal paper on self-adaptive auxiliary cube methods for multi-tenant slicing in multi-layer computing power networks contributed to this direction. Further work on integrated self-adaptive topology abstraction for carbon-efficient multi-domain computing-aware routing was progressed, supporting the development of energy- and carbon-aware routing and abstraction methods.
The second group of results concerns power-consumption modelling for virtualised Open RAN systems. The project produced models and predictive methods to estimate power use in Open RAN and virtualised radio-access-network architectures. This resulted in published conference papers on the design and analysis of power-consumption models for Open-RAN architectures and on machine-learning-based predictive models for power consumption in virtualised O-RANs. Further manuscripts were drafted on cross-platform power-consumption prediction using transfer learning, energy consumption in next-generation radio access networks, sleep-mode-aware energy modelling, energy-latency trade-offs in O-RAN with distributed baseband processing and AI inference, and energy savings through resource sharing. These results provide a technical basis for comparing O-RAN architectural choices and identifying where energy savings may be achieved.
The third group of results concerns digital-twin and machine-learning-based modelling for optical communication systems. The project developed predictive models for Raman tilt, stimulated Raman scattering spectrum, and related physical-layer behaviour in dense wavelength-division multiplexed and ROADM-based transmission systems. This resulted in published conference papers on Raman tilt prediction for digital-twin modelling of ROADM-based transmission systems and on CNN-based transfer learning for stimulated Raman scattering spectrum and tilt prediction. These results can support more efficient modelling and control of optical transmission systems by reducing dependence on repeated direct measurement or purely analytical approximation.
The project also generated experimental and testbed-oriented knowledge through work linked to OpenIreland and COSMOS. These activities strengthened the relevance of the modelling work by connecting it to programmable and experimental network infrastructures. The resulting methods and lessons can be reused in future testbed studies on energy-aware network operation, optical systems, Open RAN, wireless networks, and edge/cloud infrastructures.
The potential impacts are primarily scientific and knowledge-based. For academic researchers, the results provide models, benchmarks, and published evidence that can be compared, extended, and reused in future studies. For testbed communities, the project contributes methods and lessons for energy measurement, modelling, and validation in programmable infrastructures such as OpenIreland and COSMOS. For telecom and ICT stakeholders, the results provide evidence relevant to energy-aware deployment, virtualised Open RAN design, sustainable AI-enabled services, and future network-energy assessment methods. For policy and sustainability stakeholders, the project supports a clearer understanding of the energy implications of future digital infrastructure and the need to treat energy efficiency as a core design requirement.
ENGCoN is not positioned as an immediate commercial product or deployment-level industrial intervention. Its most realistic impact pathway is to strengthen the evidence base for future demonstrations, comparative benchmarking, network-energy assessment, follow-up research proposals, and sustainable 5G/6G and Open RAN design. The results may also support larger collaborative projects on green digital infrastructure, particularly where experimental validation, cross-platform comparison, or energy-aware network orchestration are required.
Further uptake will require several actions. First, the remaining publication pipeline should be completed so that the models and findings are available through peer-reviewed channels. Second, the methods should be further validated across heterogeneous platforms, workloads, and operating conditions, including future OpenIreland and COSMOS-based experiments where possible. Third, reusable outputs such as code, data, model descriptions, and metadata should be made available through suitable repositories when this is technically feasible and consistent with confidentiality and intellectual-property constraints. Fourth, future work should move towards larger-scale demonstrations that connect power models, network-control decisions, and real or emulated traffic conditions.
Further success would also benefit from continued international collaboration, especially across European and US testbed communities, and from engagement with industry-facing stakeholders working on Open RAN, edge/cloud infrastructure, optical networking, and AI-enabled network management. Standardisation and policy uptake would be supported by clearer benchmarking practices for reporting network energy use, carbon impact, and energy-latency trade-offs. Commercialisation is not the main route at this stage, but any protectable software, modelling workflow, or data-driven method that emerges from follow-up work should be assessed through Trinity College Dublin support structures for intellectual-property review before public release.
The project results provide reusable scientific knowledge and technical methods for understanding and reducing energy use in next-generation communication networks. Their main value lies in enabling further research, benchmarking, testbed validation, and future collaborative work on sustainable, energy-aware digital infrastructure.
The first group of results concerns energy models and benchmarking methods for future networks. The project developed methods and architectural analyses that support the comparison of energy use across network components, computing resources, and multi-domain communication systems. A published journal paper on self-adaptive auxiliary cube methods for multi-tenant slicing in multi-layer computing power networks contributed to this direction. Further work on integrated self-adaptive topology abstraction for carbon-efficient multi-domain computing-aware routing was progressed, supporting the development of energy- and carbon-aware routing and abstraction methods.
The second group of results concerns power-consumption modelling for virtualised Open RAN systems. The project produced models and predictive methods to estimate power use in Open RAN and virtualised radio-access-network architectures. This resulted in published conference papers on the design and analysis of power-consumption models for Open-RAN architectures and on machine-learning-based predictive models for power consumption in virtualised O-RANs. Further manuscripts were drafted on cross-platform power-consumption prediction using transfer learning, energy consumption in next-generation radio access networks, sleep-mode-aware energy modelling, energy-latency trade-offs in O-RAN with distributed baseband processing and AI inference, and energy savings through resource sharing. These results provide a technical basis for comparing O-RAN architectural choices and identifying where energy savings may be achieved.
The third group of results concerns digital-twin and machine-learning-based modelling for optical communication systems. The project developed predictive models for Raman tilt, stimulated Raman scattering spectrum, and related physical-layer behaviour in dense wavelength-division multiplexed and ROADM-based transmission systems. This resulted in published conference papers on Raman tilt prediction for digital-twin modelling of ROADM-based transmission systems and on CNN-based transfer learning for stimulated Raman scattering spectrum and tilt prediction. These results can support more efficient modelling and control of optical transmission systems by reducing dependence on repeated direct measurement or purely analytical approximation.
The project also generated experimental and testbed-oriented knowledge through work linked to OpenIreland and COSMOS. These activities strengthened the relevance of the modelling work by connecting it to programmable and experimental network infrastructures. The resulting methods and lessons can be reused in future testbed studies on energy-aware network operation, optical systems, Open RAN, wireless networks, and edge/cloud infrastructures.
The potential impacts are primarily scientific and knowledge-based. For academic researchers, the results provide models, benchmarks, and published evidence that can be compared, extended, and reused in future studies. For testbed communities, the project contributes methods and lessons for energy measurement, modelling, and validation in programmable infrastructures such as OpenIreland and COSMOS. For telecom and ICT stakeholders, the results provide evidence relevant to energy-aware deployment, virtualised Open RAN design, sustainable AI-enabled services, and future network-energy assessment methods. For policy and sustainability stakeholders, the project supports a clearer understanding of the energy implications of future digital infrastructure and the need to treat energy efficiency as a core design requirement.
ENGCoN is not positioned as an immediate commercial product or deployment-level industrial intervention. Its most realistic impact pathway is to strengthen the evidence base for future demonstrations, comparative benchmarking, network-energy assessment, follow-up research proposals, and sustainable 5G/6G and Open RAN design. The results may also support larger collaborative projects on green digital infrastructure, particularly where experimental validation, cross-platform comparison, or energy-aware network orchestration are required.
Further uptake will require several actions. First, the remaining publication pipeline should be completed so that the models and findings are available through peer-reviewed channels. Second, the methods should be further validated across heterogeneous platforms, workloads, and operating conditions, including future OpenIreland and COSMOS-based experiments where possible. Third, reusable outputs such as code, data, model descriptions, and metadata should be made available through suitable repositories when this is technically feasible and consistent with confidentiality and intellectual-property constraints. Fourth, future work should move towards larger-scale demonstrations that connect power models, network-control decisions, and real or emulated traffic conditions.
Further success would also benefit from continued international collaboration, especially across European and US testbed communities, and from engagement with industry-facing stakeholders working on Open RAN, edge/cloud infrastructure, optical networking, and AI-enabled network management. Standardisation and policy uptake would be supported by clearer benchmarking practices for reporting network energy use, carbon impact, and energy-latency trade-offs. Commercialisation is not the main route at this stage, but any protectable software, modelling workflow, or data-driven method that emerges from follow-up work should be assessed through Trinity College Dublin support structures for intellectual-property review before public release.
The project results provide reusable scientific knowledge and technical methods for understanding and reducing energy use in next-generation communication networks. Their main value lies in enabling further research, benchmarking, testbed validation, and future collaborative work on sustainable, energy-aware digital infrastructure.