【作者】 孙武剑；

【导师】 李鸥；

【作者基本信息】 解放军信息工程大学 ， 通信与信息系统， 2013， 博士

【摘要】 现有的无线频谱资源的基本分配完毕与已分配频谱资源的利用率不高造成了目前的频谱资源危机，也促进了一种对授权频段进行二次利用的新型技术——认知无线电（Cognitive Radio，CR）技术。认知无线电赋予认知用户/设备智能性，通过感知周围的无线环境，在不影响或者在主用户所能承受的干扰范围内，允许认知用户在时间、空间、频率等多个维度上全方位使用频谱资源。认知无线电为解决当前的频谱危机提供了较好的解决方案，受到国际学术界和产业界的普遍关注，并且被公认为是无线通信领域的“下一个大事件”。由于认知用户是在授权用户空闲时伺机利用频谱机会，所以当主用户回归时必须及时清空信道。这种授权信道可用机会的高度动态变化特性使得认知网络中的频谱资源管理面临一系列新的挑战，频谱机会的管理被认为是与频谱机会的获取具有同等重要的意义。为了能有效地发现、整理并利用感知出来的无线频谱机会，本文依托国家863探索导向类项目“基于行为预测的认知网络资源优化分配技术”，本着提高无线资源利用率的优化目标，在保证授权用户服务质量的前提下，对认知无线网络的频谱管理展开研究，研究内容包括频谱资源的感知、决策、共享和切换，主要研究成果及创新点如下：（1）提出了一种感知有效性和可靠性联合折中优化的频谱感知方法。基于聂曼-皮尔逊准则推导了感知优化模型，该模型能在充分保护主用户遭受次用户碰撞限制的前提下将频谱感知的两个关键参数即检测时间和检测周期建模为二维目标最优化问题进行联合优化，并基于能量检测建立了联合优化的数学模型，进行了理论推导和分析，给出了可使认知网络系统频谱利用率最大的检测时间和检测周期的求解方法。与主流文献中固定虚警概率最大化检测概率的优化方法不同，本方法基于聂曼-皮尔逊准则推导时采用了固定漏检概率以最小化虚警概率的原则以期较好地将次用户对主用户造成的干扰限制在指定的范围内。对于一组给定的网络参数(目标检测概率、接收信噪比、主用户行为等)，该方法都可以找到最优的检测周期和检测时间。仿真结果表明，该方法在保证检测可靠性的前提下有效地提高了频谱利用率。（2）提出了一种衰落信道上噪声功率不确定环境下的协作频谱感知方法，该方法综合考虑信道的衰落特性和噪声功率的不确定性，并采用感知性能相对较好的协作频谱感知方法以补偿信道衰落和噪声功率浮动带来的检测性能下降。针对认知网络中接收单元射频前端的复杂无线环境，将次用户接收机前端面临的热噪声、次用户之间的干扰、主用户对次用户造成的干扰等噪声综合建模为功率动态变化的随机过程，并进一步考虑到无线信道的衰落特性，推导出了噪声功率在指数分布条件下瑞利衰落信道、莱斯衰落信道和Nakagami-m衰落信道上的频谱感知检测概率的闭式表达式。为了进一步提高感知可靠性，克服无线信道衰落和噪声不确定性带来的不利影响，提出了基于D-S理论的协作频谱感知方法，使用改进的高斯概率分布函数作为D-S合并算法的基本概率赋值函数，推导出了相应协作频谱感知的判决准则。仿真结果表明，虽然噪声功率不确定带来的不利影响无法彻底消除，但协作感知还是有效地提高了频谱感知的可靠性。（3）提出了一种时延最优的负载均衡频谱决策方法。面对频谱感知收集的多个可用频谱资源（机会），认知用户需要选取某种度量准则下最优的信道接入。如果不加以控制，所有认知用户都选择此信道接入，必然会使该最优信道变得拥挤甚至造成频谱利用率的降低（热点信道效应）。本文所提的时延最优频谱决策方法将次用户业务按照信道可用度将负载均衡地分配至各个可用信道，在综合考虑了频谱的异质特性、底层的感知错误以及信道的失效特性引起的性能下降基础上，首先提出了基于概率的负载均衡频谱决策方法，并基于“抢占-恢复”式优先级排队模型展开分析，推导出了次用户驻留时间最短的可用信道资源的概率分配向量；然后，为了克服概率预测的非实时性，将瞬时的感知结果引入频谱决策，提出了一种基于感知的频谱决策方法，该方法在主用户回归时允许次用户重新感知可用信道以实现通信的连续性；最后，基于上述两种方法，提出了一种时延最优的负载均衡频谱决策方法。该方法能在均衡负载的前提下，自适应地选择基于概率还是基于感知的决策方法，以实现次用户驻留时间的最优化。仿真结果表明，相比于随机信道接入，该方法具有较好的频谱利用率和负载均衡效果。（4）提出了一种频谱覆盖共享模式下的次用户接入控制方法。根据感知的结果和决策的判断，认知用户选择出某种准则意义上的最优信道接入。实际中，由于感知错误等原因，难免会与主用户发生碰撞并造成主用户通信质量的下降。对于频谱覆盖式共享方法，一旦发生分组碰撞，主用户必须重传整个分组，所以对主用户干扰的控制就转换为对发送时机的限制。通过借鉴IEEE802.11链路层协议中分布协调功能（Distributed CoordinationFunction，DCF）的“p-坚持”接入控制思想，提出了一种保护主用户通信质量的次用户信道接入控制方法，该方法以主用户与次用户发生碰撞导致的重传从而引起的时延扩展（比）作为主用户的干扰限制，然后以次用户的平均驻留时间的最大值作为约束，并以此推导出认知用户的控制接入概率的表达式，从而避免了次用户的贪婪接入引起的主用户通信质量下降。本方法的推导并未对其它网络条件加以限制，网络模型的求解可以任意改变对主用户通信过程不同保护程度下的控制接入概率。理论分析和仿真结果表明，该方法能在保障主用户通信质量的同时兼顾了次用户通信质量，实现了较高的频谱利用率。（5）提出了一种频谱铺垫式共享模式下次用户传输时间优化方法。与前述频谱覆盖式共享优化方法不同的是，该方法对模型的假设条件进一步放宽，并且在频谱覆盖式条件下展开研究。该方法允许次用户在主用户占用信道时以小于噪声的极低功率继续发送数据，但是次用户发送数据时造成的干扰必须加以严格限制来保证主用户的通信质量（信息传输速率要求）。与传统的将干扰约束为约束次用户的发射功率来限制对主用户干扰不同的是，本方法以主用户和次用户在发送数据时的重叠时间作为干扰的度量，以主用户的最大数据传输速率（信道容量）门限值为约束条件，以最大化次用户信息发送速率为优化目标，将次用户对主用户的干扰体现在信道容量公式里的信噪比上，辅以次用户业务的队列稳定性作为服务质量的约束条件，给出了重叠时间干扰度量与传输时间的关系表达式，通过约束次用户的发送时间实现对干扰的控制。该方法将主用户的业务行为模型建模为交替更新过程，而且不要求主用户网络为时隙网络系统，前条件较为宽松，能在主用户业务行为先验条件未知的情况下得到通用的闭式解。理论分析和仿真结果表明，该方法能在干扰约束前提下提高次用户的信息发送速率。（6）提出了一种混合式频谱切换模式下的备用信道列表生成方法。次用户经过感知、决策并在共享策略下接入某个授权信道后，由于主用户可能在任意时刻重新占据信道，被打断的次用户只能重新选取授权信道以便继续传输未完成的通信过程（在碰撞造成的干扰范围内）。为了保证次用户通信的连续性和透明性，提出了时延意义上最优的次用户通信过程被中断后的备用切换信道生成方法，该方法以最短次用户在认知网络中的最短驻留时间为优化目标，将次用户在系统中的驻留时间定义为正常传输所需的时间与由于经历主用户多次打断而引起的信道切换时延之和，推导出了平均意义上次用户的一次通信过程中驻留时间的数学模型，该模型在综合考虑主用户业务统计特征、次用户的流量特性和信道的异构特性的基础上，基于“抢占-恢复”式优先级排队模型推导出了次用户平均驻留时间的拉普拉斯变换解及其相应的备选信道生成方法。进一步，该方法将无线链路的失效概率考虑到模型中，推导出了固定链路失效概率条件下次用户平均驻留时间的变换解表达式以及相应的备选信道列表生成方法。理论分析和仿真结果表明，相比于随机切换方法，该备用信道切换方法具有较好的时延性能。更多还原

【Abstract】 The total wireless spectrum available has been currently allocated to designated agency butmuch of the allocated spectrum is largely underutilized, which leads to the spectrum resourcecrisis. Cognitive radio is a key enabling technology capable of collecting unused part of thespectrum opportunity and serves as a remedy for plausible spectrum shortage problem. Cognitiveradio incorporates intelligence into cognitive users/devices through consistent monitoring ofsurrounding wireless environment, then identifies spectrum holes in the domain of time, spaceand frequency etc., and allows cognitive user to access in an opportunistic way. All theprocedures aforementioned should takes interference to primary users into consideration,threshold of interference caused by cognitive user should be met and ideally no interference at allis desirable. As a promising solution for current spectrum crisis, cognitive radio has been widelyacknowledged and closely followed by academic institutions and industrial agencies.Internationally, cognitive radio is generally regarded as a “Next Big Thing” for the possibilityand feasibility of solving the spectrum-related problems nowadays.Since cognitive user opportunistically exploits spectrum bands when primary users areabsent, immediate vacancy from temporarily-accessed channels should be performed as soon asthe primary user returns. Spectrum opportunity fluctuation, however, is hard to capture due to therandom behavior of primary service traffic, which poses many brand-new challenges forspectrum management in cognitive radio networks. In supporting and finishing nationalhigh-tech exploration-oriented project-‘Optimization techniques of wireless resource allocationin cognitive radio networks based on behavior prediction’, this thesis investigates the spectrummanagement techniques of cognitive radio networks in terms of four successive problems,namely identifying spectrum opportunities, collecting and scheduling of identified opportunities,sharing hierarchically with primary users and vacant timely from on-transmitting channels, allpursuing at improving spectrum utilization. We identify the above four questions to fourfunctions within spectrum management framework, namely spectrum sensing, spectrum decision,spectrum sharing and spectrum mobility, all of which taking end-to-end goal of primary user intoconsideration. With maximizing in mind, we perform systematical research on the four functionsof spectrum management with constrained interference to primary users, main contributions andinnovations of this dissertation are summarized as follows:(1) An efficiency-reliability tradeoff algorithm for spectrum sensing is proposed to optimizesensing parameters such as sensing duration and sensing period. Based on Neyman-Pearsoncriterion, spectrum sensing model of maximizing spectrum utilization is formulated. Theproposed sensing model offers sufficient QoS guarantees to primary users by means of posingstrict restriction on missed detection probability. Instead of optimized a single parameter, wemodel the sensing problem in a two-dimensional manner and joint optimization of both sensing duration and sensing period is conducted. Energy-based detection method is employed to obtaindetailed mathematical analysis together with theoretical deduction and spectrum utilization isoptimized. Differing from methods adopted in most popular literature focusing on fixed falsealarm rate to maximize detection probability, we fix probability of missed detection to minimizeprobability of false alarm using Neyman-Pearson criterion, which causes no explicit harm toprimary users within the define of specified value. For a given configuration of cognitive radionetworks, i.e., target detection probability, received signal-to-noise ratio of cognitive user,primary user behavior, optimal sensing duration and sensing period can be obtained withspectrum utilization greatly improved and spectrum reliability safely guaranteed. Simulationresults reveal that the proposed tradeoff algorithm proves to be efficient with sufficientprotection to primary user.(2) A cooperative spectrum sensing method is proposed with noise power fluctuationwireless channel fading taken into consideration. Noise power uncertainty and channel fadingdegrades the performance of spectrum sensing, cooperative spectrum sensing, which gives betterreliability performance compared with local sensing, is adopted to mitigate the adverse effect ofchannel fading and noise power fluctuation. Since radio frequency front-end of secondary user isexposed to a number of interference such as thermal noise of transceiver, interference fromprimary users and peer secondary users, we model the overall effect of the aggregatedinterference as additive white Gaussian noise with average power fluctuated. Furthermore,fading characteristics of wireless channel is also incorporated into mathematical modeling andclosed-form expression of detection probability are conducted on Rayleigh channel, Ricianchannel and Nakagami-m channel with noise power exponentially distributed. To further offsetthe performance degeneration caused by noise power fluctuation and improve the reliability ofspectrum sensing, cooperative sensing based on D-S combination theory is employed. ImprovedGaussian distribution function is borrowed to serve as basic probability assignment function anddecision criterion is obtained to determine whether the licensed channel is busy or not.Simulation results show that although sensing reliability cannot be mitigated, cooperativesensing still gives better reliability performance compared with local spectrum sensing.(3) A delay-optimal spectrum decision method with traffic load balancing is proposed. Onfinding multiple available spectrum opportunity, secondary user needs to decide which channel ismost suitable to access according to specified optimization criterion. All the traffic load rushed tothe best channel, if no control measure is taken, and the best channel would soon turn out to bethe most crowded, underutilized channel (Hotspot channel effect). The proposed methodbalances aggregated traffic load to multiple spectrum opportunities according to availability ofeach channel. Sensing errors, spectrum heterogeneity and outage effect of channel areincorporated into the formulation and analysis of the problem. Firstly, a probability-basedspectrum decision is proposed to evenly distribute traffic load of secondary user and detaileddeduction based preemptive-resume priority queuing theory is performed. Probability allocationvector of spectrum opportunity is obtained to minimize residual time of secondary user. Secondly,to overcome the lack of real-time knowledge of channel state, a sensing based spectrum decision method is proposed. In order to guarantee seamless communication for secondary user, thesensing based method allows secondary user to freely select channel after collision with primaryuser occurs. Finally, based on the aforementioned two spectrum decision method, adelay-optimal spectrum decision is presented with traffic load reasonably balanced. Base stationof secondary user selects suitable decision method to realize optimization of residual time.Simulation results reveals, compared with random decision method, our proposed method givesbetter delay performance as well as load balancing effect.(4) An access control method in the spectrum-overlay sharing environment is proposed.Based on sensing results and decision policy, the most available channel is selected according tospecified optimization criteria. In practice, due to sensing errors and other underlying layerrelated problems, it is inevitable to collide with primary user, causing QoS degeneration for bothprimary user and secondary user. As for the spectrum-overlay sharing mode, once collided, thewhole packet/frame needs to be re-transmitted. So, in order to restrict collision rate, it is of greatsignificance to choose the proper sending instant of secondary user. Borrowed from IEEE802.11Distributed Coordination Function with p-persistent transmission policy, a control access methodis proposed to sufficiently protect the communication process of primary user. Subjected toextended transmission time (ratio) restriction of primary user caused by collision with secondaryuser and average residual time restriction of secondary user, access probability for secondaryuser is obtained and greedy access is prohibited to further protect the transmission of primaryuser. The construction of the optimization problem is of general purpose and no otherassumptions are demanded. Furthermore, solution-solving process of formulated problem isquite generic, arbitrary extent of protection for primary user can be set and access probability canbe thus obtained. Simulation results show that spectrum utilization is improved while QoS ofboth primary user and secondary user is guaranteed.(5) A transmission time optimization method is proposed in the spectrum-underlay sharingenvironment. Unlike spectrum-overlay sharing method mentioned above, the proposed methodfurther relaxes restrictions and conducts research in a spectrum-underlay fashion, which allowssecondary user continue sending with ultra-low transmission power, but interference caused bysecondary transmission to primary user should be strictly restricted to ensure that no explicitharmful disturbance comes into being. Differing from classical interference-controlling measurethat pushes hard restriction on transmission power of secondary user, our proposed methoddefines interference in terms of overlapped transmission time of primary user and secondary user.With guaranteed maximal data rate (channel capacity) of primary user, we aim to maximize datarate of secondary user. We convert quantified interference to signal-to-noise ratio of Shannoninformation formula within the stability condition of secondary user service queue. Relationbetween transmission time of secondary user and overlapped time interference to primary userare derived. Our proposed method models primary user behavior by alternative renewal processand no slot system requirement is demanded. With constraints relaxed and solutionclosed-formed, the proposed method still works even if no prior information of primary userbehavior were available. Theoretical analysis and simulation results verify our proposal that maximal data rate of secondary user can be achieved with maximal data rate of primary userguaranteed.(6) An optimal target channel sequence generation method is proposed in hybrid spectrumhandoff environment. Finishing sensing, decision and sharing of spectrum opportunity, spectrumaccess is permitted and multiple disruptions may occur due to the random return of primary user.Once collision occurs, the interrupted secondary transmission should immediately preserve thecontext of communication process and find another channel available to resume unfinishedtransmission (the extent primary user are exposed to secondary user’s collision is severelyrestricted). To maintain seamlessness and transparency, target channel sequence of secondaryuser experiencing disruption is researched in pursuit of minimal residual time spent in cognitivenetworks. Our proposed method attempts to shorten the average residual time composed ofnormal transmission time of each secondary user and handoff delay caused by multiple spectrumhandoffs. Mathematical modeling of average residual time of each transmission is conductedbased on preemptive-resume priority queuing theory and long-terms statistics of primary userand secondary users as well as spectrum heterogeneity are involved. For the sake ofmathematical heterogeneity, only Laplacian transform solution of average residual time isobtained but optimal target channel sequence is obtained. Furthermore, channel outageprobability is introduced to the analysis and derivation and average residual time with constantoutage probability is then obtained in a Laplacian transform expression, and correspondingoptimal target channel sequence is also obtained. Theoretical analysis and simulation resultsreveal that, compared with random handoff mechanism, our proposed handoff method offersbetter delay performance.更多还原