Energy-efficient spectrum sensing using a novel adaptive hybrid learning for CR-IoT networks

Abstract

The rapid expansion of internet of things (IoT) networks has intensified spectrum scarcity due to the massive growth in wireless device connectivity. Cognitive radio sensor networks (CRSNs) offer a promising solution by enabling dynamic access to underutilized spectrum bands. However, existing spectrum sensing techniques in CRSNs often suffer from high energy consumption, low adaptability, and limited prediction accuracy posing challenges in energy-constrained environments. This paper proposes an energy-efficient spectrum sensing (EESS) framework using an adaptive hybrid learning model (AHLM) that integrates wavelet transform-based signal decomposition (WT-SD), deep reinforcement learning (DRL), entropy-based hierarchical clustering (EHC), and meta-learning-based transfer learning (ML-TLM). WT-SD extracts key spectral features, while DRL with policy-gradient optimization dynamically predicts spectrum availability. The EHC mechanism clusters sensor nodes to minimize redundant sensing, and ML-TLM enhances adaptability with minimal retraining. The proposed model achieves substantial improvements over traditional methods. Experimental results show a 36% reduction in sensing time, 60% lower energy consumption than energy detection (ED) methods, and an 18.3% increase in network lifetime. The model also achieves a probability of detection of 0.998 and accuracy of 98.1%. These results confirm that the proposed EESS-AHLM framework provides a scalable and intelligent solution for energy-aware spectrum sensing in next-generation cognitive radio (CR)-IoT environments.