%Aigaion2 BibTeX export from HES SO Valais Publications
%Monday 31 August 2026 06:14:11 AM

@ARTICLE{,
    author = {Buzcu, Berk and Finarelli, Laura and Zeydan, Engin and Baranda, Jorge and Tzanis, Nikolaos and Kanakaris, Nikolaos and Rizzo, Gianluca},
  keywords = {6G networks, AI-native automation, energy-aware network management, multi-domain orchestration},
     title = {An Energy-Aware Multi-Domain Orchestration Architecture for Sustainable 6G Networks},
   journal = {Computer Communications},
    volume = {Selected Papers from the 21th Wireless On-demand Network systems and Services Conference},
      year = {2026},
  abstract = {Next-generation 6G infrastructures are expected to behave as intelligent, sustainable, and self-
optimizing systems that bring together heterogeneous communication and energy domains. This
paper introduces an energy-aware, multi-domain orchestration framework for 6G networks that
rests on a hierarchical control model composed of an Inter-Domain Management and Orchestration
(IDMO) layer, per-domain Management and Orchestration (DMO) entities, and Infrastructure Domain
Managers (IDMs), all coupled through a Service-Based Management Architecture (SBMA). The
framework further embeds Virtual Power Plants (VPPs) and an inter-domain Energy Management
System (EMS) that interface, both physically and logically, with the network fabric while preserving
the operational independence of local grids. The EMS processes standardized, cross-domain energy
data and forecasts into locality-aware reports for the IDMO, which drives cross-domain service
decomposition, placement, and reconfiguration. The design is realized as an executable prototype
in which the IDMO ranks candidate domains with a Multi-Criteria Decision Analysis (MCDA) score
that weighs service capability, free capacity, latency slack, and an EMS-supplied greenness signal,
and whose VPP–TSO interface solves a power flow on a CIGRE benchmark grid driven by hourly
ENTSO-E data for two real national grids. On a two-domain proof of concept it reduces carbon by
31.2\% (from 803 to 553 gCO2eq/h) at 100\% slice acceptance and within the slices’ latency SLAs,
and sustains 95\% acceptance under load where a performance-first orchestrator retains only 72\%. On
the NOBEL-EU and G{\'{E}}ANT backbones (200 requests, 40 ms SLA) it reduces carbon by 19.9\% and
16.0\% at 95.5\% and 90.0\% acceptance}
}