【作者】 王威；

【导师】 何晨；

【作者基本信息】 上海交通大学 ， 通信与信息系统， 2010， 硕士

【摘要】 本文考虑了认知无线电网络中的自私频谱感知和分配策略问题:M个自私的次用户寻找恰当的时机接入N个授权频段。受限于硬件条件,每个次用户能且仅能选择一个授权频段进行频谱感知,并根据感知结果适时竞争接入授权频段。不同的授权频段可能给次用户带来不同的效用值。出于自私性,每个次用户总是做能够带来最大效用值的频段选择。我们的目标是设计一种最优网络频谱感知策略在满足各次用户自私性前提下兼顾网络整体性能,也即最大化网络吞吐量。我们将这一问题描述为一个非合作频谱感知博弈,其稳定的频谱感知决策就对应于一个Nash均衡。我们提出了一种新颖的贪婪算法,该算法可以高效地计算出所有的纯策略Nash均衡,并且对效用函数的具体形式没有过多要求。基于该算法,我们随后提出了其改进版本,并从理论上证明了该改进算法可以计算出最优纯策略Nash均衡,即具有最大的网络吞吐量。同时,我们给出了改进算法的分布式MAC协议。大量的仿真数据和实验结果验证了理论推导的正确性,并展示了所提出的算法的优异性能。进一步的,我们证明了所提出的贪婪算法具有普适性:对所有拥有严格单调效用函数的单拥塞博弈问题均可适用。进而对此类博弈,所有的纯策略Nash均衡均可以在O(nlogm)内求解出。所提出的算法解决了该类博弈问题,因而具有一定的理论价值。更多还原

【Abstract】 In this paper, we consider a noncooperative cognitive radio network with M self-ish secondary users (SUs) opportunistically access N licensed channels. Every SUchooses one channel to sense and subsequently compete to access (based on the sens-ing outcome) to obtain the channel utility. Different channels may have different utili-ties. Each SU selfishly makes a sensing decision to maximize its obtained utility. Theobjective is to design an optimal sensing policy with maximum network throughput.This problem is formulated as a noncooperative game where a stable sensing policyreaches a Nash Equilibrium (NE). A novel greedy algorithm with great efficiency isproposed to calculate all pure-strategy NE for a large class of utility functions. Byslight modification, the algorithm is able to reach an optimal pure-strategy NE withthe maximum network throughput. The algorithm can be practically implemented asa MAC protocol in a distributed way with negligible communication overhead. In-tensive simulation experiments verify the correctness of the theorems and reveal theeffectiveness of our algorithm. Furthermore, we have shown that our algorithm canbe extended to all singleton congestion games with strict monotonic utility functions.And for these games, all NE can be calculated within O(n log m).更多还原