自适应波束形成技术作为阵列信号处理领域中一个重要的研究方向，广泛应用于雷达、声呐、无线通信、语音信号处理、医疗成像等领域。相比于数据独立的波束形成方法，自适应波束形成技术具有良好的空间分辨能力以及干扰抑制能力等。实际阵列系统中面临着各种非理想因素，对阵列信号处理波束形成技术造成不可忽视的影响，如信号传播的畸变、阵列的形变、阵元的幅度和相位起伏、回波数据统计分布的变化等等。因此，实际的阵列波束形成必须要解决不同应用环境下的适用性和稳健性问题。现有的一些稳健波束形成方法虽然一定程度上缓解了传统自适应波束形成方法在非理想条件下性能恶化问题，但是也都有各自的局限性。本文针对现有波束形成方法所面临的期望目标信号方向矢量严重失配、约束参数选择困难、回波数据统计分布变化、约束矢量失配等问题，开展非理想条件下的稳健波束形成方法研究，并将稳健方法应用于空时自适应处理。本文的主要贡献和创新点概括如下： 1、针对传统单不确定集约束的稳健波束形成方法存在的不确定集尺寸参数选择难且在方向矢量误差较大时性能损失严重的问题，提出了基于多不确定集约束的稳健波束形成方法。与传统的稳健方法不同，该方法利用多个小不确定集合去覆盖整个大的不确定区域。由于原始问题是一个非线性非凸约束的二次优化问题，提出了两种迭代方法进行求解。第一种方法将原问题重排变换至高维空间，然后利用半正定松弛技术来重构问题，最终通过循环迭代求得波束形成器的自适应权矢量。第二种方法引入多个辅助变量，然后利用半正定松弛技术将问题中的非线性项通过一阶泰勒展开近似为线性项。在期望目标信号方向矢量误差很大时，所提方法提供了很好的稳健性以及很高的输出信干噪比。 2、针对非高斯信号环境二阶统计量描述在波束形成中造成的干扰抑制剩余和输出性能损失问题，提出了基于最小离差无失真响应的稳健波束形成方法，并结合期望目标信号的方向矢量严重失配情况，通过最小化阵列输出离差的同时施加多个不确定集约束来解决此问题。由于多不确定集约束，原始问题是一个非凸的ι〓范数最小化问题，提出了一种迭代的投影梯度算法将原问题转变为非凸二次约束二次规划问题。在每一步迭代中，首先引入一个辅助变量，将问题重排转化至高维空间，然后利用半正定松弛技术重构此问题进行求解。得到的权值在每一步迭代中沿着梯度方向进行更新，并投影到问题的可行域。在非高斯信号环境下，当存在大的方向矢量误差时，相比于其它所有基于最小方差准则以及基于最小离差准则设计的方法，所提波束形成方法可显著提高系统的输出信干噪比。 3、针对机载雷达探测超低空目标时，由于超低空目标与环境之间的耦合多径效应，产生的“镜像”虚假目标严重恶化协方差矩阵估计和运动目标检测性能问题，将稳健的波束形成方法应用于空时自适应处理(STAP)，提出了基于时域优化的多点联合幅相约束的稳健空时自适应处理方法。首先给出了超低空目标环境下的信号模型，分析了超低空目标及其“镜像”目标引入的多普勒扩展特性。通过在待检测目标位置临近的空间和时间二维域内多点采用幅度和相位的联合约束，并基于保形约束设计了主瓣响应矢量，推导了空间和时间二维响应矢量的解析解，实现了STAP二维响应的主瓣保形，克服目标多径效应造成的检测性能损失。 关键词：稳健自适应波束形成，最差性能优化，方向矢量严重失配，阵列信号处理，最小离差，非高斯信号，稳健空时自适应处理，联合幅相约束

 Adaptive beamforming is an important part of array signal processing which has been widely used in many areas, such as radar, sonar, wireless communication, medical imaging and others. Compared with the data-independent beamformers, adaptive beamformers have better resolution and interference rejection capability. However, the conditions are usually nonideal in practical applications, which will lead to a performance degradation of adaptive beamformers. These nonideal conditions contain nonuniform transmit medium, distorted antenna and the unknown statistical distribution of the samples. Therefore, the beamformers must be valid and robust under different conditions. However, all of proposed robust adaptive beamformers have drawbacks. In this dissertation, focusing on the problems that the performances of many robust beamformers degrade dramatically if large steering vector mismatch occurs, and the beamformers based on the minimum variance criterion is not statistically optimal under non-Gaussian signal assumption, novel robust beamforming methods have been proposed. In addition, applying the robust beamforming methods in array signal processing into the space-time adaptive processing (STAP), a robust STAP method has been proposed. The main contributions and innovative points of this dissertation can be briefly summarized as follows: 1.Focused on the problem that traditional robust beamformers based on one uncertainty set suffer from difficulty in selecting an appropriate size of the uncertainty set, and their performances degrade dramatically if large steering vector mismatch occurs, a robust adaptive beamformer using multiple uncertainty sets has been proposed. Multiple small uncertainty sets have been utilized, instead of using a single large set, to cover the whole large uncertainty region. The original problem is nonlinear and nonconvex. To tackle this problem, two solutions have been devoted. The former method first recasts the original problem in high dimension, then the problem is tackled with semidefinite relaxation (SDR) by dropping the rank-one constraint. Finally, calculate the weight vector of the beamformer iteratively. Another method uses auxiliary variables. This method does not impose the zero constraint on the imaginary part of the array response at the desired direction. Firstly, introduce auxiliary variables and utilize the SDR technique to reformulate the problem. Then, convert the nonlinear term to a linear one by using first-order Taylor approximation. Therefore, the original nonconvex problem has been transformed to a convex one. The optimal Taylor expansion point is obtained by using iterative approach. The proposed method offers a significant performance improvement in case of large steering vector mismatch. 2.Focused on the problems that the beamformers based on the minimum variance (MV) criterion is not statistically optimal under non-Gaussian signal assumption, and large steering vector mismatch occurs, a robust beamformer based on minimum dispersion variance distortionless is proposed. This method minimizes the dispersion of the output of the array and utilizes multiple small uncertainty sets. The l〓- norm minimization problem turns to be nonconvex because of these multi-uncertainty-set constraints. To solve the problem, a projected gradient algorithm is utilized to transform the original problem into a nonconvex quadratically constrained quadratic programming problem. In each iteration, introduce an auxiliary variable firstly and recast the problem in high dimension. Then, use the semidefinite programming relaxation technique to deal with the transformed nonconvex quadratic programming problem. The proposed beamformer offers a significant performance improvement in case of large steering vector mismatch for non-Gaussian signals compared with the conventional schemes. 3.Focused on the problem that a supper-low-attitude target results in target detection performance loss or target tracking error due to the secondary interaction between the moving and environment clutter, applying the robust beamforming methods in array signal processing into STAP, a robust STAP method based on an optimized joint magnitude and phase constraints in the temporal domains is proposed. First, the general radar signal model under the supper-low-attitude target environment is presented. Then, the Doppler spread characteristics caused by the supper-low-attitude target and its false image counterpart are analysed. With joint magnitude and phase constraints imposed on and around the target, the mainlobe is well maintained and the performance loss due to Doppler spread is avoided. The robustness and detection performance is significantly improved with the proposed method. Keywords: robust adaptive beamforming, worst-case performance optimization, large steering vector mismatch, array signal processing, minimum dispersion, non-Gaussian signals, robust space-time adaptive processing, joint magnitude and phase constraint