Robust DOA Estimation for Non-Coherent Sub-Arrays with Non-Uniform Noise Variances

This paper addresses Direction-of-Arrival (DOA) estimation for non-coherent sub-arrays with non-uniform noise variances for both known and unknown cases. Conventional approaches typically assume uniform noise power across subarrays, an assumption that often does not hold in practice and may produce large estimation errors. To solve this problem, we first provide a weighted least squares (WLS) solution for the case of known noise variances. For the more challenging scenario of unknown noise variances, we further propose a novel Iterative Alternating Optimization (IAO) algorithm, whose block-coordinate descent structure efficiently solves this challenging joint estimation problem. We prove that the DOA Cramér-Rao lower bound (CRLB) is invariant to the knowledge of the noise variances and that our algorithm is guaranteed to converge. Simulations demonstrate the proposed method’s robustness and superior performance, in stark contrast to conventional approaches which suffer from large estimation errors.