Robust adaptive beamforming based on a method for steering vector estimation and interference covariance matrix reconstruction

Abstract Robustness is an important factor in adaptive beamforming because various mismatches that exist in reality will lead to considerable performance degradation. In this paper, a robust adaptive beamforming (RAB) algorithm is proposed based on a novel method for estimating the steering vectors (SVs). The interference-plus-noise covariance matrix (INCM) is then reconstructed by the SVs and their corresponding power estimates. As we know, the Capon power spectrum is actually a function of the SV defined in a high-dimensional domain, in which the actual SVs correspond to the highest peaks. The nominal SVs may lead to relatively lower power amplitude points around the peaks when mismatches exist, and these peaks are located in the directions of gradient vectors at the lower power points obtained by the nominal SVs. Therefore, to obtain the actual SVs, we first construct a subspace for each nominal SV in a small neighborhood of angles. Then we get the gradient vector, which is orthogonal to the corresponding nominal SV neighborhood, using a subspace-based method. Finally, we search along the gradient vector to obtain the adjusted SV that generates the highest Capon power amplitude. The interference covariance matrix (ICM) is reconstructed by the adjusted interference SVs and corresponding Capon power amplitudes. The actual SV of the signal of interest (SOI) is estimated as the adjusted SV of the SOI. Simulation results demonstrate that the proposed method is robust against various types of mismatch and is superior to other existing reconstruction-based beamforming algorithms.