Outlier-resistant Bayesian tensor completion for angle estimation in bistatic MIMO radar under array element failures

Conventional angle estimation algorithms for multiple-input multiple-output (MIMO) radar are susceptible to array element failures and impulsive noise, which makes achieving accurate estimates in practical applications challenging. To remedy this, we propose an outlier-resistant Bayesian tensor completion algorithm for joint direction-of-departure (DOD) and direction-of-arrival (DOA) estimation in bistatic MIMO radar under element failures and impulsive noise. First, we constructed a slice-missing tensor signal model that is corrupted by outliers. To achieve better low-rank regularization on this tensor, we convert it into a structured tensor with randomly missing entries. We then design an outlier-resistant Bayesian tensor completion model, which accounts for array element failures and the "heavy-tailed" nature of impulsive noise. In the proposed model, the reconstruction of missing entries represents array element failures, while Student-t distribution models the impulsive noise in the measurements. A variational Bayesian inference scheme is developed to address the proposed model, which alternates among estimating the factor matrices, recovering the tensor rank, and mitigating impulsive noise. Finally, the completed factor matrix is used to extract DODs and DOAs using the shift invariance technique. Simulation results confirm the outstanding performance of the proposed algorithm in estimating target numbers and angles under element failures and impulsive noise.