感知毫微微蜂巢網路中基於低複雜度合作頻譜感測之分散式頻譜存取

Abstract

於本研究中，吾人提出一種適用於感知毫微微蜂巢網路中(cognitive femtocell network)基於頻譜感測資訊之頻譜配置法。此方法可在頻譜隱匿式及頻譜重疊式之混合策略下(joint spectrum overlay and underlay strategy)有效增加頻譜重覆使用率。吾人首先將頻譜配置的問題轉化成一個非合作式賽局的形式，並且使用賽局理論中的遺憾匹配學習演算法(regret-matching learning algorithm)來獲得最佳解。為了得到在頻譜配置時需要的相關感測資訊，吾人提出一種基於最大最小特徵值(maximum-minimum eigenvalue; MME)感測技術的新式合作頻譜感測法，在提出的方法中，基地台間可利用廣播的方式與鄰近基地台交換特徵值的資訊，以達到合作的目的並增進感測效能。這種新的合作方式使得基地台在交換訊息時有著較小的傳輸負擔以及較少的運算複雜度。最後，電腦模擬結果顯示出吾人提出的新式的合作頻譜感測法可以達到和傳統的方法幾乎相同的效能，以及吾人提出的基於頻譜感測資訊之頻譜資源配置法，可在頻譜隱匿式及頻譜重疊式之混合策略下有效增加頻譜重覆使用率。

In this thesis, we develop a sensing-based spectrum allocation scheme with joint spectrum overlay and underlay strategy to increase the frequency reuse factor among cognitive femtocell networks. The spectrum allocation problem is firstly formulated as a non-cooperative game and the regret-matching learning algorithm in game theory is used to solve the problem. To obtain the sensing-based information for spectrum allocation, a new maximum-minimum eigenvalue-based (MME-based) cooperative sensing scheme is proposed which shares the information of eigenvalues with neighboring femtocells via wireless broadcasting to enhance detection probability in cognitive femtocell networks. The new cooperative scheme uses a small overhead for exchanging local information and requires a low complexity for local computations. Finally, computer simulation results show that the new spectrum sensing scheme can achieve nearly the same sensing performance as conventional methods and the proposed sensing-based spectrum allocation scheme with joint spectrum overlay and underlay strategy can increase the frequency reuse factor effectively.