針對感知無線網路設計之分散式多通道偵測策略

Abstract

為了提升頻譜使用效率，近年來已經有許多致力於感知無線電(Cognitive Radio)的研究。在分散式的感知無線電網路中由於硬體的限制，對於感知無線電使用者而言，偵測所有頻帶在實作上是不太實際的。因此，部分可觀察馬可夫決策程序(Partially Observable Markov Decision Process, POMDP)可用於部分可觀察的環境中，提供感知無線電使用者關於網路環境的充分資訊。現有以POMDP為基礎的通訊協定，為了提升頻譜使用機會以及系統效能而使用了通道聚集的技術。然而，頻譜偵測所需的時間並未被考慮，在實際的環境中當通道數目增加時，過多的偵測時間冗餘將無可避免地導致頻譜機會的損失。因此，在此論文中，在考量頻譜偵測時間冗餘的情況下，我們提出了隨機式多重頻帶偵測(Stochastic Multiple Channel Sensing, SMCS)之通訊協定，以讓感知無線電使用者能夠根據部分觀察的通道狀態，針對聚集吞吐量(Aggregated Throughput)最大化做出最佳決策。藉由我們所提出的SMCS通訊協定，感知無線電使用者能夠快速地適應多變化的環境，這是因為多重通道偵測的最佳決策是動態調整的。除此之外，通道偵測的問題也進一步延伸至不完美偵測的情境，這將大幅降低吞吐量，因為主要使用者和感知無線電使用者之間可能會發生封包碰撞。因此，為了改善碰撞問題，感知無線電使用者除了通道選擇之外，還必須決定通道偵測的時間長度。我們提出了兩階段式的隨機式多重頻帶偵測(TSMCS)之通訊協定以將感知無線電使用者的聚集吞吐量最大化並且還能達到主要使用者服務品質的要求。關於不完美偵測的問題將被證明是凸性最佳化問題，因此藉由使用迭代式次梯度法(Iterative Approach with Subgradient)，將能有效地解決此問題。模擬結果顯示我們所提出的SMCS和TSMCS通訊協定皆能有效地讓感知無線電使用者的聚集通吐量最大化。

A great amount of research has devoted to cognitive radio (CR) in recent years in order to improve spectrum efficiency. In decentralized CR networks, it is not realistic for the CR users to sense the entire spectrum in practice due to hardware limitations. Consequently, the partially observable Markov decision process (POMDP) can be utilized to provide the CR users with sufficient information in partially observable environments. Existing POMDP-based protocols adopt channel aggregation techniques in order to improve spectrum opportunities and system performance. However, the required time for channel sensing is neglected, which is considered inevitable to result in large sensing time overhead and spectrum opportunity loss in realistic environments with increased number of the channels. In this thesis, based on the partially observable channel state information in consideration of sensing overhead, the stochastic multiple channel sensing (SMCS) protocol is proposed to conduct the optimal channel selection for maximizing the aggregated throughput of the CR users. By adopting the proposed SMCS protocol, the CR users can highly accommodate themselves to the rapidly varying environment since the strategy for channel sensing is dynamically adjusted. Moreover, the channel sensing problem is further extended to the imperfect sensing scenario, which can severely degrade the throughput due to packet collision between the primary users (PUs) and the CR users. Consequently, in addition to channel selection, it is required for the CR users to determine the channel sensing time in order to address the collision problem. The two-phase SMCS (TSMCS) protocol is proposed to maximize the aggregated throughput of the CR users while still fulfilling the PUs' QoS requirements. The problem associated with imperfect sensing is proved to be a convex optimization problem and can therefore be efficiently solved by exploiting iterative approach with subgradient method. Numerical results show that the proposed SMCS and TSMCS protocols can effectively maximize the aggregated throughput for decentralized CR networks.