【作者】 史兢；

【导师】 张乃通；

【作者基本信息】 哈尔滨工业大学 ， 通信与信息系统， 2007， 博士

【摘要】 智能天线技术是近年来通信领域的研究热点,它利用自身的空分能力增加了信道容量,提高了频谱效率,扩展了覆盖范围,使系统性能得到极大的提高。同时,这些独特的优势使其成为了后第三代(Beyond 3G,简称B3G)或第四代(简称4G)移动通信的关键技术。波束形成技术源于智能天线,增强期望、抑制用户干扰是其独特的技术优势,也是空分多址(SDMA)得以实现的基础。在宽带移动通信中,干扰源数目往往大于天线阵元的个数。因此,方向图的综合优化是一个多目标多参数的非线性优化问题。遗传算法在解决这类优化问题上具有明显的优势。但传统的基于遗传算法的波束形成方法大多存在算法复杂,收敛不稳定等缺陷。因此,本文提出了一种基于遗传算法的复合波束形成方法,与传统算法相比,它具有计算量低、稳定性好、收敛速度快等特点。首先,本文介绍了智能天线技术的基本概念,分析和比较了智能天线传统波束形成算法的性能。同时介绍和比较了智能天线波达角(Direction of Arrivals,简称DOA)估计的一些实现方法。最后,介绍了遗传算法的基本概念以及实现方法。这些为以后的遗传算法和复合波束形成的结合提供了充分的理论依据。基于遗传算法的波束形成方法需要准确估计相干信号的DOA,鉴于现有相干信号DOA估计方法的不足,本文首先提出了一种新的简单自适应加权前后向空间平滑算法SWFBSS(Simple Weighted Forward Backward Spatial Smoothing),它将协方差矩阵的对角子阵前后加权平滑,实现对相干信源的解相干;其次优化了前向加权空间平滑算WFSS(Weighted Forward Spatial Smoothing)算法的计算,使其计算量可以比原来降低近一半;最后改进优化了前后向加权空间平滑算法WFBSS(Weighted Forward Backward Spatial Smoothing),通过数学推导证明了平滑后前后向等价协方差矩阵的两部分存在简单的共轭置换关系,利用这种性质可以减少近一半的计算量。这三种方法可以在低信噪比的情况下对空间相近相干信号的DOA进行有效地辨识,其中双向平滑算法还可以减少天线阵列的孔径损失。在基于遗传算法的复合波束形成方法中,需要预先知道信号的波达角及相对多径能量来完成干扰抑制分配策略。针对多用户相干源的情况,本文提出了一种估计相对多径能量的方法:采用基于累积量的空间特征算法和本文提出的加权空间平滑算法估计得到各个用户信号的空间特征向量和DOA,利用矩阵求逆法来估计各个用户相对多径能量;同时为了增强算法的稳定性,引入了最小二乘法和总体最小二乘法。基于遗传算法的波束形成在解决波束图多目标多参数的非线性优化问题上具有很大的优势。针对已有方法存在的问题,本文提出了一种基于遗传算法的复合算法,把天线阵元分割为两部分,一个子阵列用来抑制干扰,另一个用来压低旁瓣,两个子阵列波束图乘积复合为最后的波束图。理论和仿真都表明,这种复合算法具有计算量小、收敛速度快和稳定性好的特点。更多还原

【Abstract】 In recent years, smart antenna technology has become the research hot point in communication field. It takes advantage of spatial diversity ability to increase channel capacity, improve spectrum efficiency and enlarge cover area, which improve system performance largely. So these special advantages make smart antenna become the key technology of beyond 3G and 4G.Beamforming stems from smart antenna, with increasing desire signal and suppressing interference as its technical superiority, which is also the foundation for the realization of SDMA. In wideband wireless communication, the number of interference is usually larger than that of antenna array elements. Therefore, optimization of antenna array pattern synthesis is a problem of non-linear multiobjective and multiparameter optimization. Genetic Algorithms (GA) have outstanding advantages in such case. However, the traditional GA-based methods for beamforming are usually with high complexity and unstable convergence. The paper proposed a GA-based composite beamforming method. In comparison with traditional methods, the proposed method has lower computation, stronger convergence stability and faster speed.Firstly, this paper introduces elementary concepts of smart antenna, analyzes and compares the performances of conventional smart antenna beamforming algorithms. At the same time, some methods for smart antenna DOA estimation are discussed and compared. Finally, elementary concepts and implementation methods of GA are introduced. All of these provide enough theoretical basis for the combination of GA and beamforming.In GA-based beamforming methods, the DOAs of coherent singals need to be estimated accurately. This paper proposes and optimizes some DOA estimation methods of coherent singals. Firstly, we propose a new simple weighted forward-backward spatial smoothing-SWFBSS, which spatially smoothes diagonal sub-matrixes of covariance matrix by weighted forward and backward method to de-correlate the coherent sources perfectly. Secondly, we optimize WFSS (weighted forward spatial smoothing) algorithm to decrease its computational amount to half. Finally, we improve and optimize WFBSS(weighted forward backward spatial smoothing). Through mathematical deduction, we obtain the simple relationship of conjugate permutation between the two parts of covariance matrix processed by WFBSS. This property can help to reduce half of the total computational amount. These three methods can discriminate the near coherent signals under low SNR. Moreover, the forward backward spatial smoothing methods can decrease the aperture loss of antenna array.In the GA-based composite beamforming methods, we need to know the DOA of signals as well as those relative mulit-path energies to perform interference suppressing assignment scheme. For multi-user coherent sources, this paper proposes a method to estimate the relative multi-path energy. In this method, spatial eigenvectors and DOAs of signals are obtained by the spatial signature algorithm based on cumulant and the proposed weighted spatial smoothing algorithms, then the relative multi-path energy of each user are estimated by matrix inversing operation. To enhance the stability of algorithm, LS (least squares) and TLS (total least squares) are introduced.GA-based beamforming has obvious advantages in the multi-object, multi-parameter and nonlinear optimization of antenna topology. This paper proposes a GA-based composite algorithm, which divides the antenna elements into two parts, one sub-array suppressing interferences and the other lowering side lobes. The final antenna topology is obtained by multiplying antenna topologies of both sub-arrays. Both theoretical analysis and simulation show the composite algorithm has low complexity of computation, fast convergence and reliable stability.更多还原