Hierarchical Federated Learning Framework for Privacy-Enhanced RAN Optimization in Distributed 5G and Private LTE Systems

This paper presents Hierarchical FL-RAN, a novel federated learning framework for privacy-preserving Radio Access Network (RAN) optimization in distributed 5G and private LTE systems. By leveraging a multi-tier aggregation approach, local models are trained at edge RAN nodes and aggregated progressively through intermediate controllers and central servers, reducing communication overhead and enhancing scalability. The framework integrates domain-specific feature encoding with temporal filtering to capture key network KPIs such as interference patterns and handover metrics while ensuring data privacy. Simulation results demonstrate faster model convergence and improved resource efficiency compared to conventional federated learning methods. The proposed framework enables secure, real-time, and distributed intelligence for RAN optimization in heterogeneous, multi-tenant wireless networks