Leveraging Generative Adversarial Networks for Spectrum Utilisation Performance Improvement in Cognitive Radio Networks

The rapid expansion of wireless communication technologies has intensified the demand for radio spectrum, leading to spectrum scarcity and inefficiencies resulting from static allocation policies. This study introduces a novel approach to optimizing spectrum utilization in Cognitive Radio Networks (CRNs) by leveraging Generative Adversarial Networks (GANs). CRNs facilitate dynamic spectrum access, enabling secondary users to exploit underutilized frequency bands without disrupting primary users. GANs, known for generating high-fidelity synthetic data, are employed to model and predict spectrum occupancy patterns accurately. The research utilizes Nigerian spectrum data within the 47 MHz to 1 GHz range to train GAN models, producing synthetic data that mirrors real-world spectrum usage. Evaluation metrics, including accuracy, spectral efficiency, and convergence rate, indicate the model's robustness. The GAN model achieved a 92.5% accuracy rate, a spectral efficiency of 6.5 bps/Hz, and converged efficiently within 85 epochs. These results underscore the model's effectiveness in enhancing spectrum utilization and addressing data scarcity challenges. This scalable framework holds significant potential for deployment in regions with limited computational resources and pervasive spectrum scarcity. The study concludes by recommending the integration of GAN-based models into real-world CRN systems and outlines future research directions to further optimize spectrum efficiency and network performance.