High Dynamic Range Gridless DOA Estimation for Single-Snapshot Sparse Automotive MIMO Radar

Single-snapshot direction-of-arrival (DOA) estimation is fundamental in automotive MIMO radar, where fast-chirp waveforms typically provide only one measurement per range–Doppler cell. In this regime, classical subspace techniques such as MUSIC and ESPRIT suffer from rank deficiency, while grid-based sparse recovery methods are affected by basis mismatch and increased complexity for large, irregular sparse planar arrays. This paper proposes a gridless single-snapshot DOA estimator operating on the continuous steering manifold of general sparse planar virtual arrays. A coarse-to-fine matched-filter search on the physically consistent array manifold provides geometry-aware initialization. The detected components are subsequently refined via a Gauss–Newton scheme with closed-form Jacobians and embedded in a lightweight RELAX-style cyclic update to mitigate multi-target coupling. To resolve merged components in high dynamic-range scenarios, a local one-versus-two model comparison based on the Bayesian Information Criterion (BIC) is employed. Model order is determined through a statistically grounded generalized likelihood ratio test (GLRT)–based stopping rule that controls the target overestimation probability.Simulation results demonstrate robust performance under weak–strong target coexistence and reliable resolution below the nominal Rayleigh limit. ... mehrThe theoretical computational complexity scales linearly with the number of array elements and coarse grid evaluations, followed by low-dimensional Gauss–Newton refinements, making the method suitable for real-time automotive radar processing.