Performance Analysis of Two-tier Cognitive Radio Networks With K-Means Clustering in Nakagami-mFading

Sammanfattning

This thesis investigates the performance of a two-tier underlay cognitive radionetwork (UCRN) by jointly analyzing the effects of Nakagami-m fading and Kmeans-based spatial clustering. A MATLAB-based simulation framework is developed in which secondary users (SUs) are spatially distributed within the simulated area according to a homogeneous Poisson point process (PPP). The users are then grouped into clusters and transmit under strict interference constraints imposed by the primary receiver (PR). Nakagami-m fading models channel variability,and a global TDMA-based orthogonal access scheme is employed to eliminateintra-cluster and inter-cluster interference. Closed-form analytical expressions for outage probability (OP), symbol errorrate (SER), and ergodic channel capacity (CAP) are derived. Analytical results for OP and CAP are validated through Monte Carlo simulations. Although ananalytical SER expression is obtained, it involves special functions that are not supported in the installed MATLAB environment; therefore, SER performance is evaluated via simulation. The results demonstrate that increasing the Nakagami-m parameter reduces fading severity, leading to improved reliability through lower OP and SER. Spatial clustering enhances performance by reducing average transmission distances and improving interference management. Increasing the number of clusters improves system-level performance, while for a fixed clustering configuration, a slight decrease in ergodic capacity is observed as fading becomes milder under interference limited power control. Clusters located closer to the primary receiver are subject to stricter transmit power constraints under the underlay model, resulting in reduced received SNR and comparatively degraded OP, SER, and capacity. Overall, the study shows that fading characteristics, spatial clustering, and underlay power constraints jointly determine UCRN performance within the considered modeling assumptions.