Ra-Csrm: A Federated Learning-Assisted Chaotic Search Resource Mapping for Efficient Qos Aware Spectrum Allocation in Cognitive Radio Networks

Abstract

Cognitive Radio Networks (CRNs) are emerging as a dynamic solution to the increasing demand for spectrum resources. To enhance Quality of Service (QoS) in CRNs, efficient and fair resource allocation among secondary users (SUs) is imperative, especially in multi-channel environments. Conventional allocation techniques often fail to adapt to rapidly changing network conditions and do not fully address fairness and latency constraints, resulting in suboptimal QoS for SUs. A need exists for a model that ensures low latency, high allocation rates, and fairness in dynamic CRNs. This article introduces the Resource Allocation - Chaotic Search Resource Map (RA-CSRM), a novel approach integrating chaotic optimization and distributed federated learning. RA-CSRM constructs a dynamic resource map by evaluating availability among primary users (PUs) and SUs. The fairness index, updated periodically, considers QoS metrics like latency and allocation rate. Federated learning synchronizes this index across nodes. If fairness dips below the actual allocation rate, concurrent scheduling is triggered to stabilize the system. This feedback loop continues until one-to-one PU-SU mapping is achieved. Simulations show that RA-CSRM achieves an 8.15% improvement in sum rate and an 8.88% reduction in error rate under high signal-tonoise ratio (SNR) conditions. This method shows superior adaptability and efficiency over existing resource allocation techniques.