Robust algorithms for direction-finding in the presence of model errors

The application of high-resolution, subspace-based methods to narrowband direction-of-arrival (DOA) estimation relies on several critical assumptions. Two of these are that the response of the antenna array is known in all directions of interest, and that the spatial covariance of the background noise is known. Neither of these assumptions is satisfied in practice, often resulting in a serious degradation of algorithm performance. A model error sensitivity analysis is carried out for some popular high-resolution subspace-based algorithms. Theoretical expressions for the covariance of the DOA estimation error are developed and compared with that obtained by simulation. The analysis is also used to develop optimally weighted versions of the algorithms that are robust to the types of model errors considered.<>