多跳认知无线电网络动态信道接入问题研究

项目名称： 多跳认知无线电网络动态信道接入问题研究

项目编号： No.61472402

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 自动化技术、计算机技术

项目作者： 李忠诚

作者单位： 中国科学院计算技术研究所

项目金额： 83万元

中文摘要： 认知无线电技术允许次用户动态接入闲置信道，被认为是解决目前无线频谱资源稀缺与固定信道分配频率利用率低的有效方案。由于主用户活动具有未知性和信道质量处于未知且不断变化之中，次用户需要通过学习选择质量最好的信道动态接入以最大化网络的总吞吐量。目前关于认知无线电网络动态信道接入算法的研究局限于单跳网络的场景，而在实际应用中网络大多以多跳的形式部署。在单跳网络下的动态信道接入通常采用一个经典的MAB在线学习模型学习未知的信道质量，不能直接用于多跳网络，否则将引入指数级别的时间和空间复杂度，以及带来极低的吞吐性能。本项目将针对吞吐量最优的多跳认知无线电网络动态信道接入问题展开深入研究，针对多跳认知无线网络动态信道接入问题提出全新的在线学习模型用于学习未知的信道质量；然后针对随机和非随机变化这两类信道质量情况，分别设计相应的学习算法和低复杂度且吞吐量几乎最优的分布式信道接入算法。

中文关键词： 动态信道接入；认知无线电网络；多臂匪徒；在线学习

英文摘要： Cognitive radio technology, allowing secondary users dynamically access idle channels, is thought to be a promising solution to solve the dilemma of rare wireless spectrum resources and inefficient fixed spectrum allocation. As unpredictable activity of primary users and changing channel quality, secondary users have to learn for the best quality channels to maximize the total throughput. Current research on dynamic channel accessing of CRN limits to single-hop networks, whereas multi-hop networks are primary in practice. For these works for single-hop CRNs, they generally adopt the classical MAB formulation to model the online learning process on unknown channel quality. But this formulation is not applicable to multi-hop CRNs, or it causes exponential time and space complexity, as well as arbitrarily bad throughput. In this proposal we focus on the problem of throughput-optimal dynamic channel accessing in multi-hop CRNs. We first propose novel learning models to learn the unknown channel quality for dynamic channel accessing in multi-hop CRNs, and then respectively for the case of stochastic and nonstochastic channel quality, design specific learning algorithms and distributed channel accessing algorithms with almost optimal throughput in low overhead.

英文关键词： Dynamic channel access;CRN;MAB;online learning