On Dimension-reduced Adaptive Array Signal Processing And Its Application To Mimo Radar

Array signal processing finds wide applications in many domains such as radar,sonar, seismic exploration, electronic surveillance, radio astronomy, etc., and plays animportant role in both military and civilian fields. The adaptive beamforming andspace-time adaptive processing (STAP) are two main contents of array signalprocessing. This dissertation aims at the problems of adaptive beamforming and STAPin phased array radar and multiple-input multiple-output (MIMO) radar systems, as wellas optimizing antenna array geometry in MIMO radar. Theoretical analysis andcomputer simulation results demonstrate the effectiveness of the proposed methods. Themain contents of this dissertation can be summarized as follows:1. A two-dimensional adaptive beamforming (TDAB) algorithm based oncorrelation matrix for large planar phased array radar is proposed. In practicalapplication, the target signal and the interference signals of planar array are located inthree-dimensional space, which should be beamformed in the azimuth and elevationdimensions. However two-dimensional beamforming usually takes a great deal ofsampling data and has very high computation complexity by using minimum variancedistortionless response (MVDR) beamformer. In order to solve this problem,high-dimensional weight vector is written as the Kronecker product of twolow-dimensional weight vectors. By utilizing a bi-iterative algorithm, the twolow-dimensional weight vectors can be solved on the basis of correlation matrix, whichdecrease the computational complexity and the number of training samples forestimating the correlation matrix. The asymptotic convergence of the proposedalgorithm is also proved. Simulations results demonstrate that TDAB can converge verywell and achieve better performance of interference suppression than the MVDRbeamformer in the presence of short data records.2. Two two-step dimension-reduced suboptimum STAP techniques are proposed.STAP can compensate the effect of platform motion and obtain optimal performance ofclutter suppression. However, the estimation of the space-time correlation matrixrequires a sufficient number of independent and identically distributed samples, whichis more than twice the number of covariance matrix dimension. And the computationalcomplexity of calculating the inverse of covariance matrix is significantly high. Thesetwo reasons make the optimal STAP technique difficult to implement practically, andmotivate us to design suboptimum STAP methods. The firstly proposed method combines the advantages of reduced-dimension algorithm using fixed structure andadaptive reduced-dimension algorithm. Firstly, to reduce the degrees of freedom (DOFs)of the clutter, the space-time received data is pre-filtered by using the Space TimeMultiple Beam (STMB) approach. Secondly, the output data after preprocessing isprocessed by the multistage wiener filter (MWF) to optimize the weight vector, whichcan further reduce both the computational complexity and sample support requirement.The secondly proposed method firstly localizes the clutter in space-time dimension byusing the pre-filter, and decrease the DOFs of the clutter in the next step. Then theoutput data after preprocessing is adaptively filtered by utilizing the bi-iterative STAPalgorithm, and thereby further reduction in both the computational complexity andtraining requirement is achieved. Experimental results by using the measured datademonstrate the effectiveness of the proposed method.3. Two iterative three-dimensional (3D) STAP methods based on correlationmatrix for airborne radar in azimuth, elevation, and Doppler domains are proposed. Intraditional two-dimensional (2D) STAP, the elements in planar array are combined intoa linear array by using microwave synthesis, and thus the elevation is ignored. In fact,the clutter has a relationship with elevation, and the synthesis patterns in elevation ofeach subarray are different because of the unavoidable array errors, so the3D STAP canachieve better performance of clutter suppression than2D STAP. In the firstly proposedmethod, the quadratic cost function used in the optimum3D STAP is converted into twoquadratic functions by using submatrices of the space-time correlation matrix. Byiteratively optimizing two lower dimensional weight vectors in two quadratic functions,the proposed method can greatly decrease the computational load and training samplerequirement. For the airborne radar employing a large array, the quadratic cost functionused in the optimum STAP is converted into three quadratic functions, and the fulldimension weight vector can be separated into three lower dimensional weight vectors.By iteratively optimizing these lower dimensional weight vectors, the proposed methodcan significantly decrease the computational load and training sample requirement,which is easier for application.4. The beamforming method for planar phased array radar is developed into MIMOradar system. A reduecd-dimension adaptive beamformer for MIMO radar is proposed.Because of exploiting the DOFs of transmitting array, MIMO radar can achieve betterperformance in angular resolution and clutter suppression. However, the sampling datato estimate the correlation matrix and computational complexity to calculate its inverseare increased greatly because of the high-dimension of receive data. The proposed method separates the optimal weight vector with submatrices of the correlation matrix.The sample requirement and complexity are reduced by estimating the submatrx of thecorrelation matrix and calculating its inverse. In addition, combining the STAPtechnique with MIMO radar can further suppress the clutter, while the problems of highsample requirement and complexity become worse. Thus, a reduced-dimension STAPmethod for clutter suppression in airborne MIMO radar system is developed.Experimental results using simulated data in systems with different DOFs demonstratethe effectiveness of the proposed method.5. Several synthesis and optimization algorithms of sparse antenna arrays inMIMO radar are proposed. Sparse antenna arrays have advantages in fewer elements forcomparable beamwidth, lower cost and higher DOFs, while sparse arrays with randomlyspaced elements result in grating lobe effect. The optimization of array pattern synthesisin MIMO radar needs to consider the effect of the transmitting and receiving arrays,which means the antenna geometries of the transmitting and receiving arrays need to beoptimized simultaneously. In addition, MIMO radar can achieve better performance inangular resolution and clutter suppression if the maximum DOFs of system areemployed. Therefore, a new method for antenna array geometry optimization based onmultiple genetic algorithm (GA) for MIMO radar where sparse arrays are used isproposed. By introducing distance perturbations, both the transmitting and receivingantenna array geometries are modified twice by the multiple genetic algorithm, while,the problem of pattern synthesis in orthogonal MIMO radar with lower relative sidelobelevel is also solved. Simulation results indicate the validity of the introduced methods.In order to avoid falling into local optimum and prevent the premature convergence ofGA, we combine GA with Tabu search and simulated annealing which are local searchtechniuqes and propose two multiple hybrid genetic algorithms.