感知網路合作式頻譜感測之分群方法

Abstract

感知無線電 (Cognitive radio) 是近年來解決頻譜短缺問題的新興技術。其中頻譜感測在實現感知無線電系統中扮演著不可或缺的角色，因為好的感測能力可以幫助開發頻譜洞也可以降低對主要使用者 (Primary user) 的干擾。研究上很感興趣的合作式頻譜感測 (Cooperative spectrum sensing, CSS) 被認為是相對於單一使用者而言可以達到更好的感測能力，然而，當考慮回報錯誤時合作式頻譜感測會變差。另一方面，大量的次要使用者 (Secondary user) 也會使得系統回報需要的頻寬不足。分群架構的合作式頻譜感測 (Cluster-based CSS) 被提出來解決這個問題，其將次要使用者分成數群，每群中選擇一個最好的使用者做回報的架構可以得到使用者分集 (Multiuser diversity gain) 的效益。在本論文中，我們提出三種分群演算法以提升感知網路的效能，並且我們也衡量了在不同匯整規則 (Fusion rule) 下的效能。數值模擬顯示OR規則適用網路有少數的次要使用者，而AND規則在次要使用者多時有較好的效能，這也提供我們一個系統開發的方針。

Cognitive Radio (CR) is a promising technology to solve the spectrum scarcity problem in the recent years. Spectrum sensing plays an essential role for realizing the CR systems because better sensing performance can help to discover the spectrum opportunities and reduce the interference with the primary users (PUs). Cooperative Spectrum Sensing (CSS) is studied with great interest to achieve better performance than a single secondary user (SU). However, while considering the reporting errors, the performance of CSS is degraded. On the other hand, a large number of SUs will bring about a shortage of reporting bandwidth. A cluster-based CSS scheme which leads to multiuser diversity gain is addressed to solve this problem by partitioning SUs into clusters and selecting the best user to report. In this thesis, we propose three clustering algorithms to improve the throughput of CR networks and we also evaluate the performance of different fusion rules. Numerical results show OR rule is suitable when number of SUs is small while AND rule can have better performance if number of SUs is large, which provides a guideline in system development.