【作者】 潘峮；

【导师】 常永宇；

【作者基本信息】 北京邮电大学 ， 通信与信息系统， 2011， 博士

【摘要】 传统的无线通信当中,固定分配频段的方法使得频谱资源日益稀缺,然而调查显示目前的授权频段中频谱的使用并不充分。认知无线电技术因其能够以伺机接入方式使用已授权给主用户的却在某时刻空闲的频段,成为了解决上述矛盾的有效方法。本文重点研究了认知无线电中重要的频谱感知和频谱共享技术。应用频谱感知技术,可以使认知无线认知其无线电环境、了解主用户的工作状态,这项技术是认知无线电可以实现的前提。频谱共享技术实现并优化了认知系统和主用户系统的共存,这是认知无线电最终的目标。我们的主要研究点包含以下几个方面。首先,通过建模和仿真我们介绍了通过传统的接收机操作特征来衡量单用户及协作感知方式下的频谱感知性能。通过研究发现,频谱感知的灵敏度和频谱共享的效率可以通过平均检测时长和空闲信道利用率两个指标来衡量。这两个指标是有联系的,它们不仅和接收操作特征相关,而且受到认知系统时序结构的影响,需要联合考虑进行优化。我们研究的主要目标是：一方面在认知系统维持一定频谱感知时间比例的前提下,通过调整频谱感知的时长来缩短平均检测时长；另一方面是限定平均检测时长的情况下,最大化空闲信道利用率。后者是更为实际的方案。为了实现优化的目标,我们设计了在系统中调整频谱感知周期和频谱感知时间的较为实用的方法,并对这种方法进行了仿真分析。结果显示在保证认知用户频谱感知灵敏度的前提下,应用我们提出的方法能够使空闲信道利用率提升10个百分点左右。其次,协作感知虽然克服了单用户感知中隐藏终端效应及深衰落信道的问题,但是如何在协作的节点之间交换感知信息,是协作感知面临的实际问题。协作感知中,传统的硬合并方式虽然在感知性能上不及软合并方式,但是它却只需要在节点间交换很少量的信息。为了在感知性能和感知信息传输上寻找一个折中的方法,我们提出了一种分布式的变长量化方法来量化感知信息。这种方法的核心思想是让感知性能良好的认知用户有更多机会将信息精确地回传给中心节点,而那些性能较差的认知用户回传少量的信息甚至不进行回传。为了不增加下行开销,回传的信息量由认知用户自行决定。具体操作上看,首先通过简化了的认知用户频谱感知对数似然比结果,将原本在中心节点处的集中计算,分布在认知用户本地完成。认知用户根据对其结果的信心将感知信息量化成不同的长度回传给中心节点,系统总的开销量则在中心节点处通过下行广播来控制。为了进一步缩减开销量,我们设计了周期性差分量化的方法。通过仿真验证的结果表明,应用了这种新的方法后认知系统的感知性能接近理想的软合并方法,但平均开销量仅为1-2个比特。再次,要保证协作感知的可靠性,必须为这仅为数个比特的开销信息寻找一个良好的信道。然而认知系统本身没有稳定的授权频带,如果直接通过占用主用户频带来传输感知信息,不仅其本身可靠性很难保证,还有可能会对主用户产生强干扰。通过研究发现,未来可被认知系统伺机接入的主用户系统多为非扩频的宽带系统。因此我们设想可利用扩频通信方法将低速率的感知信息以较低的功率谱密度扩展到较宽的频段,再通过空间隔离来实现和主用户的频谱共享。我们以WRAN系统与DTV主用户系统之间以及认知系统与LTE主用户系统之间的共存为例仿真验证了我们的想法。分析和仿真结果表明无论是当WRAN系统位于DTV覆盖边缘,还是认知系统位于LTE小区交接处时,主用户系统的有效传输和经过扩频处理的认知系统感知开销信息传输之间的互干扰都在可容忍的范围内。最后,我们将多天线引入到认知无线电当中,来增强频谱共享和频谱感知技术。我们根据主用户和认知系统发射机之间的信道状况来进行多天线模式的选择,从而在对主用户产生有限干扰的前提下实现认知系统下行吞吐量的优化。我们还根据多天线接收信号之间的相关性,提出了利用接收到多路并行信号的自相关矩阵奇异值分解的方法来进行频谱感知。这种方法将信号能量集中在一个奇异值中,而噪声能量则分散在所有奇异值中,从而较多天线下能量检测提高了准确度。更多还原

【Abstract】 Spectrum bands are allocated in a fixed manner for the traditional wireless communications, and the available bands are becoming rare. However, the recent survey shows a low spectral efficiency in licensed band. The spectral holes in licensed band can be sensed and used to transmit information by employing cognitive radio (CR) technology. As a result, CR becomes a key to deal with this dilemma. This thesis mainly focuses on the spectrum sensing and spectrum sharing technologies in CR. CR users recognize their radio environments and detect primary user by spectrum sensing technology which is a precondition for deployment of CR technology. And the CR users realize coexistence with primary users by spectrum sharing technology which is the ultimate objective of CR. Our major works include the following.Firstly, we model the spectrum sensing and simulate the receiver operating characteristics (ROC) of both single user and cooperative spectrum sensing. In order to evaluate the performance of spectrum sensing and spectrum sharing, two important metrids are introduced. They are average detection time (ADT) and idle channel utilization ratio (ICUR). These two metrics are correlated and both depend on ROC and the time schedule of cognitive system, so joint optimization is needed. We focus on how to minimize ADT by adjusting spectrum sensing time duration while a certain fixed proportion of system time is allocated to spectrum sensing. Another more realistic research is to maximize the ICUR of spectrum sharing while the ADT is constrained. We also design a more practical way to adjust the sensing cycle and sensing time to achieve optimization. The simulation results show that the ICUR can be raised by 10 percents by using our proposed scheme while maintaining the agility of cognitive system.Secondly, although cooperative sensing has solved the problem of "hiding terminal" and deep fading channel, how to exchange the sensing information among multiple CR users is still difficult in practical deployment. Hard combination schemes underperform soft combination ones in sensing accuracy, yet with much less sensing information needed to be transmitted. A novel distributed variable quantization scheme is proposed to find a trade-off between the sensing performance and the transmission of sensing information. The essence of this proposed scheme is that CR users with more confidence send more precise sensing information back to fusion center while less confident ones send less information even nothing, and the specific amount of information is set by the CR users themselves to maintain downlink overhead at a low level. In detailed, the centralized calculation of sensing log-likelihood ratio equation at the fusion center is simplified and split into several distributed calculations which can be achieved at local CR users. Then the CR users quantize the calculation results into bits with different length based on their confidence, at the same time, the system overall overhead is controlled by the quantization factor broadcasted by the fusion center. In order to further minimize the overhead, the circular differential quantization method is devised. The simulation results show that our proposed scheme has similar sensing performance as optimal soft combination scheme, but the average overhead only costs one to two bits.Thirdly, the reliable cooperative sensing must have stable channel to exchange the sensing information with only a few bits. However, CR system itself has no licensed band. If it directly occupies the primary band, the CR and the primary users may strongly interfere each other. It seems that the primary systems whose band CR may access are largely broadband non-spectrum-spread ones. Therefore, we assume to introduce spectrum spread technology into CR system, which spread the low rate overhead to broader bandwidth with lower power spectral density. Additionally, spatial isolation is also applied to realize spectrum sharing between primary users and CR system. Our idea is validated through two practical examples. The first one is the coexistence scenario between Wireless Regional Area Network (WRAN) and DTV, and the second one is the one between a normal CR system and LTE. The WRAN is at the coverage edge of DTV, and the CR system is also at the cell edge of LTE. The primary transmission and the spread CR overhead can well coexist with each other under acceptable low interference.Finally, the multi-antenna technology is introduced to enhance the spectrum sharing and spectrum sensing. We find a way to maximize the downlink throughput by selecting a better MIMO mode based on the channel condition between CR transmitter and primary users, while the interference to primary users is constrained. The correlation among the received signals at multiple antennas is utilized to employ spectrum sensing. The proposed spectrum sensing scheme calculate the singular value decomposition (SVD) of auto-correlation matrix of parallel received signal flows. This method concentrates the signal power on one singular value, yet the noise power is averagely distributed in all singular values. Therefore, its sensing accuracy outperforms energy sensing with multi-antenna.更多还原