在無線鏈結上，利用以馬可夫決策為基礎的功率分配方式以達到有彈性的傳輸

Abstract

隨著現在行動通訊越來越普及，我們在行動設備上所使用的應用也越來越多。因此電力的消耗速度變快，而行動設備電池的壽命也越變越短。所以我們想要找出一個可節省電力消耗、但又可以有不錯傳輸表現的傳輸策略。我們想用馬可夫決策的原理，在時變有記憶性的通道當中找出一個可有力的傳輸策略。因為我們無法直接地得知現在通道的狀態，所以我們將手上的問題模組化成一個無法完整得知通道資訊的馬可夫決策程序。而這個我們找出的傳輸策略，要決定傳輸的時機，以及此次傳輸所需要用的功率等級。我們就可以在系統的表現以及電力的消耗上，達到一個平衡點。與過去有關功率分配的研究比較，最主要的不同就是我們加上了一個Worst Polling的觀念。我們在原本的策略疊代程序上面，加上了一個Worst Polling Detection，用來找出一個好的決策。這個找出來的決策能夠在最壞情況發生的前提之下，還能讓傳輸保持好的表現。也就是所謂的Flexible Communication。

With the growth of the popularity of mobile communication, the applications used in the mobile device are more than before. Thus, the speed of energy consumption is higher and the life of the battery of the device is shorter. For these reason, we want to find a policy that can achieves the goal of energy saved and good performance of transmission. We are interested in using the Markov decision theory to find a robust policy for wireless communication over time-varying channels with memory. Because we cannot directly know the channel state, we model our problem in hand as the Markov decision process based on partial or incomplete channel state information. This robust policy determines the timing of transmission and the power level at it. Therefore, we can achieve an appropriate balance between performance and energy consumption. The main different from other power management research is that we add a worst polling concept. We add a worst-polling detection to join the well-known policy iterative procedure to find a good policy so that we shall make sure the good performance of the resultant policy in the worst case in the wireless communication system. That is what we called “flexible communication”.