Research project

COALESCE-6G

Sustainability Telecoms infrastructure Spain
COALESCE-6G Header

COALESCE-6G aims to establish an innovative cloud-telco experimental platform that leverages 6G technologies to enable AI-based collaborative orchestration and resource management in multi-cloud and multi-domain infrastructures, aligning with the European Commission’s 3C Networks concept.

The project focuses on addressing key challenges, such as scalability, interoperability, and energy efficiency. By integrating advanced Radio Access Network (RAN) capabilities with edge and cloud computing resources, COALESCE-6G will deliver innovative solutions for dynamic resource allocation, Quality of Service (QoS) optimisation, and multi-domain federation.

Estimated impact

The main objective of COALESCE-6G is to lay the foundations for 6G network architecture, enabling intelligent and efficient orchestration of network resources through the seamless integration of federated edge and cloud infrastructures and the implementation of distributed AI. More specifically, the project will address the following objectives:

  • Design and implementation of advanced orchestration mechanisms enabled by Artificial Intelligence (AI) for dynamic management of network resources.
  • Federation of edge and cloud infrastructures.
  • Development of open and standardised APIs that enable the exposure and consumption of network capabilities.
  • Significantly improve energy efficiency by optimising the use of RAN and edge-cloud resources.
  • Addressing the security and privacy challenges inherent in the implementation of distributed AI systems.

In addition, COALESCE-6G will implement an advanced experimental infrastructure, interconnecting multi-site test beds, such as the i2CAT facilities in Barcelona and the 5TONIC test bed in Madrid. This platform will enable the validation of advanced 6G use cases in real environments, such as AI-based energy-efficient orchestration and the deployment of multi-domain services. The project also aligns with the vision of 6G as a ‘network of networks’ by exploring the seamless integration of distributed AI between centralised and edge nodes.

Dates 01/09/2025 - 31/08/2028
Research and Innovation Areas Artificial Intelligence (AI)
Cybersecurity
Smart Networks and Services: 5G/6G
Space Communications
Sectors Sustainability Telecoms infrastructure

Project publications

Topological Forwarding for LEO Constellations: A Scope-Specific Routing Policy in a Recursive User-Plane Architecture

Sergio Giménez-Antón, Eduard Grasa, Jordi Perelló, Maryan Rizinski, Mohammadreza Mosahebfard, John Day, Lubomir Chitkushev, Topological forwarding for LEO constellations: A scope-specific routing policy in a recursive user-plane architecture, Computer Networks, 2026, 112923, ISSN 1389-1286, https://doi.org/10.1016/j.comnet.2026.112923. (https://www.sciencedirect.com/science/article/pii/S1389128626009357) Abstract: A Low Earth Orbit (LEO) mega-constellation never stops moving, yet Internet Protocol (IP)-based routing ties each satellite's forwarding state to constellation size and recomputes routes as the shell turns. We treat routing instead as a policy of one layer of the 6G Recursive User Plane Architecture (6G-RUPA), in which a satellite's address states its position in the shell, so that forwarding becomes arithmetic: a distance policy prices the remaining path in kilometers from seven shell constants and the clock, a progress guard keeps every walk loop-free, and exception entries repair what failures break. We test the scheme in LEOPath, an open-source simulator, on four constellations and eleven shells from published filings, with Ring, +Grid, open-seam and three-laser brick-wall wirings and fifteen failure conditions, against link-state and explicit-path routing and a hop-only ablation. Per-satellite forwarding state, a proxy for router memory, falls from 351-1584 routes to the node degree, and where a full link-state table rewrites 12 to 65 routes per satellite and minute as a failure-free shell moves, a topological satellite rewrites none. Under the same addresses, forwarding matches the mean path length and delay of link-state without failures and, with northbound addresses, its delivery under failures, on paths up to 13% longer. At realistic failure rates the busiest satellite holds 25 to 474 exact exception entries, which group into 3 to 8 address regions. The price moves to the edge, as station renumbering and 7-22 ms of extra delay for attachment addressing, which a second address per station roughly halves. Whether smaller tables save lookup energy remains untested. Keywords: Routing; LEO satellites; 6G-RUPA; Satellite networks; Topological routing; Network scalability; Network architecture

A Knowledge-Based Multi-Agent Framework for Security Control Recommendation

C. Fernández-Martínez, S. Siddiqui and V. Daza, A knowledge-based multi-agent framework for security control recommendation, Knowledge-Based Systems (2026), doi: https://doi.org/10.1016/j.knosys.2026.116558