Federated Learning-based Unicast/Multicast Service Delivery over 6G O-RAN Framework

The path toward the envisioned International Mobile Telecommunications (IMT)-2030 requires seamless terrestrial and non-terrestrial networks (TNs-NTNs) convergence and shared unicast/multicast capability to face the critical 6G research verticals. The Multicast/Broadcast Services (MBS) paradigm over three-dimensional (3D) heterogeneous networks adds new degrees of freedom during coverage planning and service delivery. Embedding these technologies into an Open Radio Access Network (RAN), such as the softwarized and disaggregated architecture promoted by the O-RAN Alliance, enables native intelligent solutions with extreme flexibility. Hence, we propose a solution for shared unicast/multicast service delivery over a TN-airborne connectivity in the 6G O-RAN architecture. We present a dynamic TN-NTN selection and slice allocation algorithm based on Federated Double Deep Q-Network (FDDQN) inserted into a novel O-RAN scenario. The proposal is validated through link-level simulations, evaluating diverse network conditions and service constraints, several concurrent users, and the impact of slicing resource utilization.