Research On Array Adaptive Beamforming And Space-time Adaptive Processing Algorithms

In recent years, with the development of array antenna in the area of radar, communication,sonar and so on, the scale of the antenna arrays and the dimension of signals become larger andlarger. However, due to the limitations of work environment, hardware, manufacturing processesand other factors, the problems including computational complexity, samples demand andperformance under the influence of non-ideal factors of adaptive array signal processing algorithmsare particularly highlighted. In this thesis, on the basis of Multiple-Input Multiple-Output（MIMO）radar, airborne radar and airborne MIMO radar, adaptive beamforming and space-time adaptiveprocessing（STAP）methods are studied from two aspects of reducing dimension and robustness.The main contributions of this thesis are summarized as follows:1. Aiming at collocated MIMO radar, a robust adaptive beamforming method is developed forMIMO radar in the presence of unknown mismatches. Explicit models of uncertainties in bothtransmitted and received signal steering vectors are considered. Combined the bi-iterative algorithmand the second convex optimization algorithm, a robust adaptive beamforming method for MIMOradar is proposed. It is shown that the robust adaptive beamforming problem can be solved byminimizing a convex quadratic cost function based on the optimization of worst-case performancewhen full DOFs of MIMO radar are used. Whereas, we reformulate the quadratic cost function intoa bi-quadratic cost function by adopting a separable form for the weight vector and the minimumpoint of the new cost function can be efficiently found by combining bi-iterative algorithm (BIA)with second-order cone programming (SOCP). Compared with the robust adaptive beamformingalgorithms using full DoFs, the proposed beamformer has lower computational complexity andfaster convergence rate, while, at the same time, it provides better robustness in the non-ideal casesand reduces the training samples required.2. A model with matrix form of ground clutter data is established. Then taking full advantage ofthe prior information, such as platform velocity, radar parameters and so on, we propose a space-timeblock cancelle（rSTBC）to suppress the ground clutter for airborne radar. A least-squares cost functionassociated with the filter coefficient matrix of the STBC is established. Since the coefficient matrix isonly determined by the prior information, the proposed STBC belongs to a non-adaptive processorand owns small computation load and non-convergence process. It is shown that the proposed STBCcan also be used as an efficient pre-filtering tool before the conventional moving target indication（MTI）processing and the classical reduced-dimension adaptive processing. Moreover, since theclutter model has taken into account of the drift angle, the STBC is applicable not only to the sidelooking airborne radar but also to the non-sidelooking airborne radar.3. By utilizing the low rank property of optimal STAP weight matrix, a reduced-dimensionSTAP algorithm is introduced. It is shown that the optimum weight matrix can be naturallyexpressed as the sum of several paired spatial and temporal weight vectors. After truncated andstack processing, the weight vectors can be obtained by optimizing a bi-quadratic cost functionbased on the non-orthogonal basis iterative（NOBI）algorithm. Compared with the optimal STAPmethod, high-dimensional covariance matrix inversion is avoided in our method, and therefore, thecomputational complexity and the demand of training samples is significantly reduced.4. Extending the spatial-temporal adaptive processing (2D-STAP) to the azimuthal, elevationaland temporal three-dimensional adaptive processing (3D-STAP), two kinds of reduced-dimensionadaptive algorithms are proposed for airborne radar to suppress ground clutter. The first methodbased on the frame of the mDT-SAP applies a Doppler pre-filter to realize the first step ofreduced-dimension. Then followed by azimuthal and elevational adaptive beamforming instead oftwo-dimensional spatial beamforming, the dimensions of the processor are further reduced. Insecond method, the optimal weight vector is approximatively denoted by the Kronecker product ofthree low-dimensional weight vectors. Then we cyclically optimize a low-dimensional weightvector by applying a reduced-dimension matrix constructed by the other two fixed weight vectors.It is shown that the proposed above two methods seek weight vectors in lower dimensional spaceand perform better in small samples, thus both of them convergence fast and own low computationload and small training data demand.5. An airborne MIMO radar clutter model is established, and its characteristics are analyzedfrom the distribution of space-time power spectrum and freedoms. The problems of several classicalreduced-dimension STAP algorithms used in airborne MIMO radar are explored. Due to the factthat the received clutter data of airborne MIMO radar own a three-dimensional (transmitting,receiving and temporal) structure, which is similar to3D-STAP, then the3D-STAP methods forairborne array radar are introduced into the airborne MIMO radar space-time adaptive processing(MIMO-STAP), and their performance are testified via simulations results.