在感知無線電中繼網路下，合作式頻寬分享(Cooperative Spectrum Sharing)是一種被提出來解決頻譜使用效率不彰問題的方法。在合作式頻寬分享中，主要使用者可以藉由次要使用者中繼傳送來達到更高的傳輸率，作為回報，主要使用者必須釋放自己的部分頻寬給次要使用者進行傳送。如何促進合作與資源的分配是合作式頻寬分享的重要議題。 與大部份的文獻不同，本研究中所考慮的主要使用者與次要使用者節點都為能量採集結點，節點中電池的能量可能會或多或少甚至沒有，並非源源不絕的，因此，本論文以綠能通訊的觀點，考慮在一組主要使用者及次要使用者在合作式頻寬分享下，設計出一傳輸策略能使整體效益最大。 本論文中，吾人使用馬可夫決策過程(Markov Decision Process)來解決在能量採集、通道及電池狀態會隨著時間變動的情況下，在各種狀態決定出最佳動作，使得系統整體的傳輸中斷機率最小。再者吾人也考慮若沒有狀態轉移機率的環境中，使用Q-Learning 來決定最佳策略。最後再比較各種傳輸策略與本研究提出的方法的效能差異。 ;Cooperative Spectrum Sharing (CSS) is proposed as a solution for the inefficient utilization of the spectrum. In CSS of cognitive-radio networks, primary licensed user (PU) with poor channel condition (between its transmitter and receiver) can achieve a higher data rate by using a secondary unlicensed user (SU) as a relay. In return, the primary user possibly decides to lease part of its spectrum resource to secondary users in exchange for appropriate remuneration. The key point of CSS is how the PU and SU agree on the cooperative transmissions and resource allocations. In previous schemes, they are propose on the transmission policy with unconstrained energy. However, we focus on wireless sensor networks with energy harvesting and propose a method between PU and SU ‘s transmits policies in CSS. In our study, we formulate the problem as a Markov Decision Process (MDP) framework by which the channel states, battery conditions and harvested process are stochastic. Moreover, we also consider the condition without information of transition probability. In this case, we use the Q-learning algorithm to find the optimal transmission policies .Finally, we compare the performance of our scheme with others.
