認知無線網路：賽局模型與統計學習法之自我組織演算法

Abstract

為因應頻寬上的需求及使用效率，認知無線 (cognitive radio)技術已被視為一可行且有效的解決方法，該技術能解決目前頻寬使用上的一大問題，也就是頻寬使用的分散性，統計發現一般用戶並不會持續地使用該頻寬，造成頻寬上的頻寬洞 (spectrum hole)，頻寬使用率大幅降低，認知無線技術將認知無線網路建構在已經存在的主網路上，偵測並且使用主網路的頻寬洞，本論文主要研究認知無線網路的其中一個架構─認知無線網路通路 (CR network access architecture)，該架構主要由一頻寬經濟人 (spectrum broker)蒐集認知無線網路的資訊，進而分配頻寬資源給各個認知無線網路的認知無線基地台 (CR base station)，認知無線基地台的用戶能從這些基地台使用主網路多餘的頻寬，我們將該架構以一經濟學模型─古諾模型 (Cournot game model)來描述，認知無線基地台對頻寬經濟人提出頻寬資源需求，但同時也必須考慮競爭者，也就是其他的認知無線基地台，去評估需求大小，我們使用統計學習 (stochastic learning)演算法，調整每個認知無線基地台對頻寬經濟人所提出的頻寬資源需求，此演算法的好處是認知無線基地台只需要自我回饋的資訊即可調整頻寬需求大小，不需其他認知無線基地台或是通道環境等資訊，本論文證明該演算法能夠使系統收斂至納許均衡點 (Nash equilibrium)，實驗的結果除了反映證明外，也顯示了該演算法在整體效用的表現上具有顯著的成效。

Due to the high demand of spectrum utilization, cognitive radio (CR) network has been a promising solution to the problem of spectrum scarcity by using dynamic spectrum access technique. The CR networks is applied to the original network (or primary network) without modifying the original network. In this paper, we studied one of the CR network architectures, CR network access architecture, where the CR base stations (CRBSs) demand spectrum resources from the primary network and distribute them to the CR users. We applied an economical Cournot Game model to the system where the CRBSs are the players and compete for better performance in this game. In order to optimize the game, we proposed a stochastic learning (SL) based scheme for the CRBSs to adjust the demand amount of resources based only on the action-reward history, which means there is no need for a centralized controller.We proved that the SL-based algorithm leads the system to converge toward a Nash Equilibrium (NE) point. Numerical results correspond to the proof. The results also show that the system performs well in terms of the total utility comparing with other schemes.