在本篇論文中，主要針對由中繼網路(Relay network)及能量採集網路(Energy harvesting network)架構這兩項現今熱門主題結合所發想之合作式通訊系統傳輸策略的設計及模擬。有別於傳統通訊模式，兩通訊節點間之資料傳輸，不再受限於不好的通道狀態，可另外找尋一中繼節點幫忙轉遞資料。此外，以往的通訊節點必須透過佈置電力網或定期更換電池，來供給傳送所需之能量，而隨著綠能通訊(Green communication)的崛起及電池硬體設備的提升，使通訊節點能從環境中進行能量採集的概念也被提出。因此，於本研究中考慮二具備能量採集功能之發射節點在任意能量採集狀態、電池狀態及通道狀態的情況下，決定是否進行合作式通訊。 在此，我們將各狀態量化後，採用馬可夫決策過程(Markov decision process, MDP)和Q-learning方法，藉由調整和設計各種參數，分析如何最佳化系統的效能，並找出在各狀態下的最佳決策策略。另外，我們發現透過傳送所需單位能量大小的設計，能使系統多樣性階數上升，更能促進系統進行合作式通訊。 總結本研究主要探討目標為—考慮能從環境中收集能量的使用者，在自身通道狀態不好時，可以透過和另外一使用者進行合作式通訊來增進雙方最大效益。 ;With the rising of green communication and the upgrading of battery technology, using energy harvesting (EH) nodes as cooperative relays is a promising solution to perpetually power-up wireless sensor networks. In this thesis, we investigate the optimal transmission policies for decode-and-forward (DF) relay networks in which the two-user pairs capable of energy harvesting and a user which has sufficient energy for transmission serving as a relay to the other user whose channel condition is worse. In order to maximize the long-term reward of the system and find out the optimal policies, we formulated this problem as a discounted Markov decision process (MDP) and Q-learning framework. Through designing and adjusting the parameters, we discover that the outage probability is inversely proportional to the solar cell size and battery size. In addition, the numerical results provide that the system tends to do non-cooperation when the signal-to-noise ratio (SNR) approaches to infinity, and at last it also indicate that we could improve the diversity orders to obtain better performance and stimulate the cooperative communication via the reducing the size of energy quantum size.
