TREES aims to reduce the carbon footprint of 6G networks by integrating distributed federated learning (DFL), as a tool for predicting orchestration actions and improving energy efficiency. DFL is an Artificial Intelligence (AI) paradigm, one of whose advantages is that it consumes less energy. To achieve this goal, TREES will (i) design a new architecture and algorithms for DFL to limit energy consumption; (ii) propose methods for pooling data and learning between several applications, taking advantage of the data partitioning offered by federated learning; (iii) develop network orchestration algorithms and AI
functions to minimize the carbon footprint of deployed applications; (iv) set up, on an experimental environment, an autonomous network administration loop integrating the various tools developed in the project and real-world data to evaluate two use cases: “Leveraging Smart Power Grid for Telco” and “Energy-aware Multi-Tenant AI Function Orchestration”.
Contract nb: ANR-24-IAS1-0004