在此篇論文裡，吾人考慮了以綠能為觀點的感知無線電網路，沒有頻譜認證的次要使用者需利用頻譜檢測技術去檢測主要使用者是否正在傳輸資料，如果主要使用者是處於閒置狀態，次要使用者會接取頻譜去傳送資料。由於吾人考慮的次要使用者為一能量採集點，在每一時刻來的能量為一隨機過程，因此不能隨意地耗費能量。當次要使用者決定執行頻譜檢測，雖然消耗越多的能量，主要使用者狀態越明確，但用來傳送資料的能量與時間相對地減少。所以吾人以部分觀察馬可夫決策過程(POMDP)來找出最佳傳送策略，此決策過程會根據過往的觀察與現今的狀態來選出最好的動作。由於考慮的狀態總數過大，無法找出最佳解，所以吾人以兩種次佳解決策過程來解決上述所說的問題。最後吾人所提出的次佳解比其他次佳解方法還好。;In this paper, we consider energy harvesting cognitive Radio. No licensed, low priority secondary user needs to estimate licensed, high priority primary user’s activity by using spectrum sensing. If the sensing result is idle, secondary user will access bandwidth to transmit its own data. Due to secondary user is an energy harvesting sensor node, harvested energy is a stochastic process at any given time. Therefore we can’t waste any energy causally. When primary user decides to execute the action of spectrum sensing, the more energy for sensing is more reliable, SU has less time and energy to transmit signal. We formulate the optimal power management as partially observable Markov decision process(POMDP). POMDP will find an optimal policy based on previous observations and current informations. Thanks to the total number of states are too large, we can’t find an optimal solution. We use two sub-optimal solution to solve original POMDP. Finally, the sub-optimal solution we propose is better than another approach in other paper.
