Fast maximum likelihood DOA estimation in the two-target case with applications to automotive radar

Abstract

Direction-of-arrival (DOA) estimation of two targets using a single snapshot plays an important role in automotive radar for advanced driver assistance systems. Conventional Fourier methods have a limited resolution and generally yield biased estimates. Subspace methods involve a numerically complex eigendecomposition and require multiple snapshots or a suboptimal pre-processing for reliable estimation. We therefore consider the maximum likelihood (ML) DOA estimator, which is applicable with a single snapshot and shows good statistical properties. To reduce the computational burden, we propose a grid search procedure with a simplified calculation of the objective function. The required projection operators are pre-calculated off-line and stored. To save storage space and computations, we further propose a rotational shift of the field-of-view such that the relevant angular sector, which has to be evaluated, is delimited and centered with respect to broadside. The final estimates are obtained using a quadratic interpolation. The developed method is demonstrated with an example. Simulations are designed to assess the performance of the considered ML estimator with grid search and interpolation, and to compare it among selected representative methods. We further present results obtained with experimental data from a typical application in automotive radar.