【作者】 周敏；

【导师】 王少尉；

【作者基本信息】 南京大学 ， 电子与通信工程（专业学位）， 2021， 硕士

【摘要】 由于无线网络的快速发展和人们对更高吞吐量需求的不断增加,无线通信系统的频谱效率亟待提高。动态频谱接入(Dynamic Spectrum Access,DSA)是缓解频谱稀缺问题和提高频谱利用效率的潜在方案,它允许次级用户(即非授权用户)在无主用户(即授权用户)活动的特定时间和地点机会性地利用频谱资源。频谱感知是部署动态频谱接入系统最重要的组成部分。考虑到硬件限制,次级用户一次只能感知部分频谱或者一个信道,为了减少感知时延,本文研究了多个运营商频段间的频谱感知次序问题以及主动感知模式下多信道间的信道选择策略。本论文的主要贡献概括如下:1.研究了动态频谱接入系统中多个运营商频段间的频谱感知次序问题,提出了折旧汤普森采样和满意汤普森采样算法进行求解。在该问题中,我们的目标是缩短感知时长并获得尽可能多的空闲信道,从而提高系统吞吐量。考虑到频谱空闲概率的先验信息不可知,将优化问题规约成多摇臂赌博机问题。此外,还考虑了频谱空闲概率随时间变化等更符合实际的场景。仿真结果表明所提算法能够获得和最优决策相近的可用授权信道,并表现出比其他算法更强的追踪时变空闲概率的性能。2.研究了主动感知模式下多个信道间的信道选择问题,并基于现有的贝叶斯分配算法提出了Top-Two概率采样和Top-Two汤普森采样算法。我们的目标是在数据包到达之前,设计合理的信道选择策略,使得次级用户以尽可能小的感知开销充分学习未知的信道空闲概率信息,快速确定空闲概率最大的信道。在数据包到达时,将所学信息用作先验信息,从而减少搜索空闲信道的时间,缩短次级用户数据包传输时延。将优化目标看作一个“纯探索”问题。随着感知次数的增加,认为其他信道最优的后验概率收敛到0。理论结果证明了所提算法以指数速度收敛,并且相应的指数是可能达到的最优值。数值结果也表明所提算法能够以更小的频谱感知开销有效地识别最佳信道。总的来说,本论文研究了动态频谱接入系统中的频谱感知问题,重点讨论了频谱感知次序和信道选择问题,并提出了有效的算法进行求解,实现了有效的频谱感知。数值结果表明,本文所提算法能够快速找到空闲信道,减少了频谱感知时间,提高了系统吞吐量和频谱利用率,为未来动态频谱接入系统在频谱感知方面的优化提供了思路。更多还原

【Abstract】 Due to the rapid development of wireless networks and an increasing demand for higher throughput,there is an urgent need to improve the spectral efficiency of the wireless systems.Dynamic spectrum access(DSA)has been envisioned as a promising solution to alleviate the spectrum scarcity problem as well as to increase the efficiency of spectrum utilization,which allows secondary users(referred to as unlicensed users)to use the spectrum opportunistically at a particular time and location when and where primary users(referred to as licensed users)are not active.And spectrum sensing is the most important component for the establishment of dynamic spectrum access.Due to the hardware limitation,the secondary user can only sense a small portion of the spectrum band or even one channel at a time.In order to reduce the delay in spectrum sensing,we study the spectrum sensing order problem across multiple service providers and the channel selection strategy among multiple channels in proactive sensing.The main contributions of this thesis are summarized as follows:1.We study the spectrum sensing order problem across multiple service providers in the dynamic spectrum access system and propose discounted Thompson sampling and satisficing Thompson sampling algorithms to solve the problem.The goal is to reduce the sensing delay and acquire more idle channels so that the system throughput will be improved.Considering the prior information of the probability of the spectrum being idle cannot be known in advance,the optimization problem is formulated as a multi-armed bandit problem.Moreover,some practical scenarios have been taken into consideration where the probability varies at temporal scale.Numerical results indicate that the proposed methods yield idle channels almost as many as the optimal decision in hindsight and can track the time-varying probability more quickly than other algorithms.2.We consider the channel selection problem among multiple channels in proactive sensing and propose top-two probability sampling and top-two Thompson sampling based on the existing Bayesian allocation rules.The objective is to design a reasonable channel selection strategy so that the secondary user can learn more about the unknown availability probability of each channel and definitively identify the optimal channel quickly after a small number of spectrum sensing trials.Whenever the secondary user has a packet to transmit,the information that has been learned can be exploited to reduce the time searching for an idle channel and an end-to-end packet delay will be minimized.We formulate the problem as a ’pure exploration’ problem.As the spectrum sensing trails goes on,the posterior probability assigned to the event that some other channels is optimal converges to zero.Theoretical results prove that our proposed methods converges at an exponential rate,and the corresponding exponent seems to be the best possible.Numerical results show that the proposed algorithms confidently identify the optimal channel with less spectrum sensing overhead.In summary,we investigate the spectrum sensing issues in dynamic spectrum access system.In particular,we focus on the spectrum sensing order and channel selection problem,and propose some efficient methods to solve the problem,which enable efficient spectrum sensing.Numerical results indicate that the proposed methods can locate the idle channel quickly and reduce the time for spectrum sensing,thus improving the system throughput and spectrum utilization efficiency,which shed lights in the optimization of spectrum sensing in the future dynamic spectrum access system.更多还原