感知無線電設計的主要目的為增進頻譜的使用效率，因此如何有效挑選通道，便成為一個很重要的議題。一些相關研究提出了不同的方法，有依據通道剩餘時間、考量偵測錯誤機率、碰撞機率等方式，達到更好的頻譜使用率。但是這些已提出的方法卻沒有考慮到通道狀況的影響，Secondary User(SU) 的傳輸速率會受到通道狀況很大的影響，例如距離、多重路徑、遮蔽衰落等狀況。因此本論文將通道狀況納入考量，提出一個有系統的通道挑選方法。 本論文中， Fusion Center(FC) 被賦予收集 SU 偵測的資訊，藉由 FC所收集的資訊挑選適合 SU 的通道，在此提出兩類的挑選策略，集中式與分散式挑選策略。集中式挑選策略中，又分成兩種方法，考慮最大化個人傳輸速率與考慮系統吞吐量。分散式有三種通道選擇策略，包括無預先換手機制、考慮 PU 將出現之剩餘時間的預先換手機制，和考慮 PU 出現前是否能傳送完 Buffer 內資料的預先換手機制。藉由模擬結果顯示，所提出的方法皆比隨機挑選方法有更好的傳輸效能。 The purpose of Cognitive Radio (CR) technology is to improve the spectrum utilization. Several literatures proposed different approaches, which include the spectrum sensing, detection of the appearance of primary user (PU), collision detection, the channel selection, etc., to improve CR performance. Among them, the effective channel selection is more important if there are many secondary users (SU) that contend limited spectrum holes. Furthermore, as the SU may have different channel condition in each channel and this issue shall also be carefully considered during channel selection for better transmission performance. Thus, the transmission rates of SUs will be greatly affected due to their channel condition, which may be caused by the environment distance, multipath, fading and so on. This thesis proposes a systematic method for channel selection by taking the channel condition into consideration. In this thesis, the Fusion Center (FC) is equipped to collect information sensed by SUs. The proposed schemes assist SU to choose the appropriate channel by applying the collected information by FU. Both of the centralized and distributed channel selection strategies are proposed. The centralized channel selection strategy is further divided into two approaches: One is based on the maximization of individual transmission rate; and the other considers the maximization of system throughput. The distributed approach has three options for channel selection, including handoff scheme, pre-handoff scheme with the remaining time to the appearance of PU and pre-handoff scheme with the remaining time to clear the buffer before the appearance of PU. The experimental results show that the proposed methods present better performance when compare to that of the random selection scheme.
