A Federated Graph Learning Framework for Autonomous Resource Optimization in Cognitive Radio Networks

Abstract

The Cognitive Federated Graph Learning (CFGL) framework is introduced to jointly optimise three interdependent resources in cognitive radio networks: (i) frequency channel allocation for secondary-user transmissions, (ii) routing paths for multi-hop packet traversal, and (iii) transmit-power budgets as determined by per-link signal-to-interference-plus-noise ratio (SINR). CFGL incorporates federated learning to enable privacy-preserving distributed training across cognitive nodes without exchanging raw sensing data, and utilises graph neural routing to model topological dependencies for dynamic packet forwarding. A joint spectral-spatial state representation captures both frequency-domain characteristics and spatial propagation patterns. The framework applies adaptive weighting for hierarchical aggregation, informed by channel conditions and local data distributions. We provide detailed mathematical formulations for the federated optimisation objective, graph convolution operations, and the integrated routing-resource allocation problem. Experimental results indicate that CFGL achieves 18–67% higher spectral efficiency, 27–62% lower routing latency, and near-optimal fairness relative to centralised and federated baselines, while maintaining robust convergence under non-independent and identically distributed (non-IID) data conditions.