INTELLIGENT COGNITIVE RADIO DESIGN: PERFORMANCE ANALYSIS AND ML-BASED OPTIMIZATION OF ENERGY DETECTION SCHEMES

Abstract

This research presents a detailed performance analysis of a Cognitive Radio Network (CRN) employing energy detection-based spectrum sensing under realistic wireless channel conditions. The system model comprises a primary user (PU), a secondary user transmitter (SU-Tx), and a secondary user receiver (SU-Rx), where the SU-Tx senses the PU’s channel using energy detection and transmits to the SU-Rx if the channel is found idle. The communication links are affected by Rayleigh fading and distance-dependent path loss, providing a practical representation of wireless propagation. The total time frame is divided into sensing and transmission phases, and key performance metrics including detection probability, false alarm probability, missed detection probability, and average throughput are analytically derived using Gaussian approximation under the central limit theorem. The detection probability is modeled as a function of the sensing threshold, sample size, noise power, and received signal-to-noise ratio (SNR). To overcome the limitations of static threshold settings, the study integrates machine learning techniques to dynamically optimize system parameters. Reinforcement learning method is used for predicting optimal sensing thresholds and to adaptively select transmission power and sensing duration to maximize throughput. Simulation results demonstrate that adaptive threshold selection using RL significantly improves detection accuracy and throughput compared to fixed threshold methods. The study also shows how learning-based approaches reduce missed detections and false alarms, making the system more reliable and efficient in dynamic spectral environments. These findings support the development of intelligent and robust CRNs.