多输入多输出(MIMO–Multiple-Input Multiple-Output)雷达是近年来国内外研究的一个热点，其基本思想是在发射端各个发射单元同时发射不相关或正交的信号，在接收端通过匹配滤波分离出各个发射单元的信号后进行更灵活处理，以提高雷达性能。目前MIMO雷达主要分为两大类：小尺度相干MIMO雷达，其收发阵列规模小（称为小尺度），各个收（发）阵元到目标视线近似平行，目标相对于收（发）阵列具有相同的波达方向，与传统的相控阵雷达相比，这类雷达能提供更多的系统自由度，进而提高目标的可辨识度及参数估计性能；另一类是大尺度非相干MIMO雷达，其收发阵列规模足够大（称为大尺度），观测到目标的RCS不同，从而获得接收和发射两方面的空间分集增益，且在信噪比较高时，可使闪烁目标的检测性能有较大的提高。本论文主要研究了相干处理MIMO雷达参数估计方面的问题，主要工作总结如下：1. 围绕MIMO雷达利用发射孔径自由度提高雷达参数估计精度的思想，研究了MIMO雷达的工作机理。针对单基MIMO雷达阵列流形存在冗余以及同时利用收发孔径自由度导致计算复杂度增大的问题，提出了去冗余ESPRIT类降维处理方法和基于二维快速傅里叶变换(2D-FFT, Fast Fourier Transform)的降维波达方向-多普勒联合估计方法。去冗余ESPRIT类降维处理方法避免了角度搜索和相位解缠绕，在一定程度上提高了波达方向和多普勒频率的估计精度。基于2D-FFT的的数据分块方法对每个接收阵元的接收数据分块作2D-FFT，将各个分块的频域数据相干积累，找到峰值点，然后在各个分块找到峰值点对应的数据构成频域快拍，利用该降维后的频域数据实现目标波达方向和多普勒频率的联合估计。该方法具有较高的分辨率，且不需要对阵元级接收数据进行协方差矩阵的估计及其特征分解，运算复杂度低，尤其当阵元数和快拍数都较多时，优越性更加显著。2. 针对双基MIMO雷达，提出了在接收端同时实现发射角度(DOD, Direction of Departure)和接收角度(DOA, Direction of Arrival)联合估计的几种方法，包括利用发射阵或接收阵单元形成旋转因子实现了DOD和DOA自动配对的无需阵列平移的ESPRIT类联合估计方法、多天线发射两天线接收的联合MUSIC-ESPRIT方法和DOD-DOA联合估计分维处理方法，实现了低复杂度的多维参数估计。其中基于2阶矩和基于4阶累积量的MUSIC-ESPRIT方法能有效地实现白噪声和色噪声背景下的多目标角度估计。在接收端，通过单天线的MUSIC和双天线的ESPRIT算法分别估计目标的DOD和DOA，将两维参数估计转化为两个一维形式，降低了运算量和计算复杂度。3. 研究了长期困扰理论上高性能阵列技术应用的阵列误差校正与阵列处理稳健性问题。针对与信号波达方向无关的幅相误差问题，提出了双基地MIMO雷达多目标角度和幅相误差联合估计方法。该方法利用联合对角化估计含误差的接收和发射阵列流形，并通过最小化相位误差的平方和，分别得到DOD和DOA的估计，再利用估计得到的角度，通过幅相误差自校正算法得到幅相误差估计的闭式解。该方法不需要幅相误差的先验知识，对误差具有一定的稳健性。针对存在互耦的情况，提出了双基地MIMO雷达互耦自校正方法。该方法利用等距线阵下互耦矩阵的Toeplitz特性和联合收发导向矢量Kronecker直积的性质，将两维角度搜索转化为两个一维搜索，从而降低了运算量，再根据估计得到的角度可得等效的收发互耦系数，并通过奇异值分解估计发射阵列和接收阵列的互耦系数。4. 针对机载MIMO雷达目标参数估计性能受到谱展宽的地物杂波严重影响的问题，研究了基于最大似然的目标波达方向和多普勒频率联合估计方法。该方法直接利用接收数据的最大似然函数估计目标参数，但涉及两维搜索，运算量大。为了降低最大似然的计算复杂度，提出了基于二维离散傅里叶变换(2D-DFT, Discrete Fourier Transform)的波达方向-多普勒频率联合估计方法和基于空时自适应单脉冲近似实现的波达方向-多普勒频率估计方法，显著降低了运算量。

By emitting arbitrary waveforms (usually noncorrelated or orthogonal) via each transmit sensor and exploiting a matched filterbank to extract these waveforms from the reflected signals to achieve flexibility of processing and improved performance, multiple-input multiple-output (MIMO) radar has been a hot research topic. To date MIMO radar can be classified as two broad kinds. The first kind is small-scaled coherent MIMO radar, of which the antennas are closely spaced (so-called small-scaled) and therefore both the direction-of-arrival (DOA) and direction-of-departure (DOD) are identical for all the receive sensors and all the transmit ones, respectively. It has potential advantages over its counterpart (the traditional phased radar), such as increased degree-of-freedoms (DOFs), extended target identification and improved performance of parameters estimation and so on. The other kind is large-scaled noncoherent MIMO radar with widely separated antennas, namely large-scaled array. Since different antennas detect different aspects of targets, it can provide transmit/receive spatial diversity and improve the detection performance of target scintillations, especially in the case that signal-to-noise ratio (SNR) is relatively high. This dissertation mainly studies parameters estimation of the coherent MIMO radar, and they are concluded as follows:1. Around the idea that parameters estimation performance of MIMO radar can be improved by exploiting the transmit array aperture DOFs, the principle of MIM radar is studied. In colocated MIMO radar, there are redundant terms in the array manifold. Furthermore, owing to the incorporation of the transmit DOFs in receive site, the computational complexity is high. To alleviate the computational complexity problem, two algorithms are proposed. The first one is reduced-dimensional ESPRIT-like algorithm with improved estimation accuracy of angle and Doppler frequency, in which both angle searches and phase unwrapping can be avoided. The other algorithm is the 2D-FFT based reduced-dimensional angle-Doppler estimation algorithm. In this algorithm, the received signal matrix of each receiver is first partitioned into four blocks. Then 2D-FFT is applied to each block in order to achieve coherent integration. By utilizing the data corresponding to the peaks of coherent integration in each block, a reduced-dimensional data vector is constructed for angle and Doppler frequency estimation. Since the full-dimensional covariance matrix estimation and eigendecomposition are avoided, the computational cost of the presented algorithm is relatively low. The superiority is much more notable for either large array or large number of snapshots.2. Joint DOD and DOA estimation algorithms at the receive site are presented for bistatic MIMO radar, such as ESPRIT-like algorithms by means of rotational factor via either the transmit array or the receive array, combined MUSIC with ESPRIT algorithm for angle estimation by utilizing multi-antenna transmit and two-antenna receive, and the reduced-dimensional direction finding approach, which are utilized to implement multi-dimensional parameters estimation with low complexity. By employing MUSIC with single-sensor owing to its utilization of transmit array aperture and ESPRIT with two-sensor via the rational invariance property of the signal-subspace at the receive end, the DODs and DOAs of the targets can be estimated separately and paired automatically. Two-order moment and four-order cumulant based MUSIC-ESPRIT are suggested for the spatial Gaussian white noise and the colored noise, respectively.3. MIMO array error calibration and robust MIMO array signal processing which seriously influence the application of high-performance array techniques are studied.Aiming at the DOA-independent gain-phase error, an algorithm for multitarget joint angle and gain-phase error estimation for bistatic MIMO radar is proposed. Both the transmit array manifold and receive array manifold involving the gain-phase errors are obtained by joint-diagonalization. Then by minimizing the mean square of phase error for transmitter and receiver respectively, the DODs and DOAs of targets can be estimated. Finally, combined with the obtained angles, the closed-form solution of the gain-phase error can be solved by the gain-phase error self-calibration algorithm. With no need for the prior knowledge of the gain-phase error, the proposed method is much more robust. In the presence of mutual coupling, an algorithm for mutual coupling self-calibration is given. Based on the special structure of the coupling matrix of uniform linear array (ULA), the angles can be estimated directly by double one-dimensional searches without the knowledge of the mutual coupling matrices. Then the mutual coupling coefficients of the transmit array and the receive array can be solved in closed-form by utilizing the obtained DODs and DOAs, respectively.4. Since parameters estimation performance degrades seriously owing to the presence of the Doppler-extended clutter in airborne MIMO radar, an algorithm for joint DOA and Doppler frequency estimation based on the maximum likelihood principle is studied. However, it is computationally expensive as it requires two-dimensional searches. Three methods are introduced to reduce its computational complexity, including the alternative algorithm for DOA and Doppler frequency estimation, joint estimation of DOA and Doppler frequency based on 2D-FFT and the space-time adaptive monopulse approach.