Study On Robust Adaptive Beamforming Algorithm In Complex Circumstances

As one of the vital techniques, adaptive beamforming is used for enhancing a desired signal while suppressing noise and interference at the output of an array of sensors. In recent decades, adaptive beamforming has been widely used in radar, sonar, wireless communications, medical imaging, and other areas.When adaptive arrays are applied to practical problems, the performances of adaptive beamforming algorithms are known to degrade severely in the presence of even slight mismatches of such type that can easily occur in practical situations as a consequence of look direction and signal pointing errors.Especially, in scenarios of high SNR, the desired signal is suppressed as an interference signal.Therefore, it has been a popular topic to enhance the robustness of adaptive beamforming algorithms for many years.In this thesis, we analyze the factors that may affect the robustness of these methods in practical applications, and discuss some conventional robust adaptive beamforming algorithms systemically. Having the applications of robust adaptive beamforming algorithms in mind, we have proposed some effective and implementable robust adaptive beamforming algorithms.These algorithms reduce the influence of the mismatches and provide an improved robustness against the mismatches, which are summarized as follows:In scenarios of uncertainty in source direction of arrival (DOA),a novel robust constrained constant modulus algorithm (CMA) is proposed by using a Bayesian approach. The optimal weight vector of the proposed algorithm is derived from a priori knowledge about the source DOA and estimated signal steering vector from data. The proposed algorithm can provide excellent robustness for the uncertainty of DOA of the actual signal and make the mean output array signal-to-interference-plus-noise ratio (SINR) consistently close to the optimal one. Secondly, we develop robust adaptive algorithm with additional the constraint of DOA uncertainty and obtain the updating weight vector. The proposed algorithm provides an improved robustness to uncertainty in DOA under poor conditions.Moreover, the proposed robust adaptive beamforming algorithm can be computed at a comparable cost with other algorithms by using recursive method to obtain inverse matrix.In the presence of the signal steering vector mismatches and the small training sample size, least mean squares (LMS)algorithm has a low convergence speed, degrades in the output performance and even occurs instability. In order to overcome the shortages, a Bayesian approach to robust constrained-LMS algorithm is proposed in this thesis, which balances the use of observed data and a priori knowledge about the source DOA.The proposed algorithm can estimate the actual signal steering vector by using prior knowledge.The proposed algorithm reduces the influence of the mismatches, provides an improved robustness and enhances output SINR. Simulation results demonstrate that the performance of the robust constrained LMS algorithm is better than that of the traditional constrained LMS algorithm.The degradation performance of conventional adaptive beamforming algorithms may become even more pronounced in the presence of slight mismatches between the actual and presumed signal steering vectors.In this thesis, we propose robust adaptive beamforming based on the optimization of worst-case performance, which involves minimization of a quadratic function subject to infinitely many nonconvex quadratic constraints. To account for mismatches, based on linear constrained CMA, we develop robust constrained CMA algorithm.The closed-form solution is derived from optimizing cost function with nonlinear constraints.It turns out that this problem can be formulated as a convex second-order cone (SOC) program.The robust constrained CMA overcomes the problem of interference capture in CMA algorithm,provides excellent robustness against the signal steering vector mismatches and improves array output SINR consistently.Several existing algorithms are sensitive to some types of steering vector mismatches (e.g., mismatches due to imperfect array calibration, the small training sample size and so on).In this thesis, we can also consider a general form of robust adaptive beamforming and obtain significant cost function. Based on this cost function, we propose a novel neural network approach to robust adaptive beamforming. The proposed algorithm is based on explicit modeling of mismatches and a three-layer radial basis function neural network (RBFNN).In the proposed algorithm, the computation of the optimum weight vector is viewed as a mapping problem, which can be modeled by using a RBFNN trained with input/output pairs. Moreover, with LMS updating, a novel approach to robust adaptive beamforming based on worst-case performance optimization is proposed, which has a low computational complexity and provides excellent robustness against some types of mismatches.The diagonal loading factor and convergence performance is analyzed, and the convergence scope of robust constrained LMS algorithm is also given.The diagonal loading method is a simple and efficient method for improving the robustness of adaptive beamforming.However, the main shortcoming of the method is that it is not clear how to obtain the optimal value of the diagonal loading factor based on the known level of uncertainty of the signal steering vector. We develop robust adaptive beamforming algorithm based on the variable diagonal loading. To improve robustness, the output SINR is minimized to obtain the signal steering vector with additional nonlinear constraints,and the parameter in the optimal solution can be solved accurately. To decrease computation complexity, the weight vector based on the variable diagonal loading is obtained by recursive method and Taylor series.In practical situations where complete knowledge of signal characteristics is not available and it is also time-vary environment, we propose robust adaptive beamforming based on gradient descent method, which belongs to the variable diagonal loading techniques. To avoid the contradiction of convergence rate and steady errors, the variable step size is obtained by building a function between the step size and input signals.The proposed algorithm reduces the influence of mismatches, offers faster convergence rate, has an improved steady errors, and provides a nearly optimal performance in the presence of mismatches.In this thesis, based on explicit modeling of mismatches in the desired signal array response, we propose robust adaptive beamforming algorithm for implementing a quadratic inequality constraint with recursive method updating. The quadratic constraint is added to error between the assumed beampattern and actual beampattern over a small spatial region near the array steering direction, which may make the array less sensitive to the steering vector mismatches. It shows that the proposed algorithm to the case of uncertain signal steering vector belongs to the class of diagonal loading approaches, but a variable diagonal loading term of our algorithm can be optimally chosen, which is incorporated at each step. The proposed robust adaptive beamforming can be efficiently computed at a comparable cost with the conventional adaptive beamforming algorithms.Our proposed robust recursive algorithm provides an improved robustness against the signal steering vector mismatches, enhances the array system performance under random perturbations in sensor parameters, and makes the mean output array SINR consistently close to the optimal one.