AI-Driven Spectrum Sensing for Cognitive Radio Networks in Dynamic IoT Environments

Abstract

Because of the quick rise in IoT devices and the extra traffic on wireless networks, the spectrum management system we have is being challenged, so efficient and intelligent use of the spectrum is now needed. CRNs have stepped forward as a successful approach because they allow secondary users to make use of licenced frequencies that primary users are not using at the time, so there is no interference. Still, orthodox methods for sensing the spectrum such as energy detection, matched filtering and cyclostationary analysis, may not work well in environments that are fast-changing, have lots of noise or involve different types of WiFi users, so they can miss some users, report false alarms more often and leave unused portions of the spectrum. To overcome the issues mentioned, the paper introduces a new framework using both DRL and CNNs to make spectrum sensing more context-sensitive and adaptive in IoT-enabled CRNs. The CNN model is created to identify features in the spectrograms of radio frequencies and represent the changing density of the signals. These features are provided to a DRL agent which increases its sensing policy through interacting with a simulated network, all while considering wireless metrics such as SNR, the status of energy, movement speed and levels of interference. Different IoT scenarios are set up such as static and mobile nodes, different SNRs and heavy interference and the hybrid model is examined carefully in each case. The simulations show that there is a big improvement of 25–40% in detection and spectrum efficiency as well as faster convergence to optimal policies, compared to the existing methods. In addition, the system works well when channel conditions change, demonstrating that it is capable and practical for future wireless networks. These results prove that AI technology will improve CRN performance and help meet the varied needs of tomorrow’s connected world.