Optimization of High-resolution and Ambiguity-free Sparse Planar Array Geometry for Automotive MIMO Radar

Abstract

The next-generation 4D imaging automotive radar is characterized by high angular resolution, unambiguous detection, low latency, low cost, and small size. This study provides an enhanced analysis of the angular ambiguity function (AAF) for planar MIMO arrays, and pioneers a method for a more accurate evaluation of angular resolution using the main lobe width (MLW). Then the 2D expanded beam pattern (EBP) is introduced to assess the field-of-view (FOV), region of interest (ROI), sidelobe level (SLL), and normalized resolution intuitively and precisely. After constructing the sophisticated 2D element spacing and aperture constraints for planar MIMO arrays, the optimization of array geometry is creatively formulated as a novel Domino sparse optimization problem aiming to minimize the MLW while sufficiently suppressing the SLL, which is inspired by the sequential fall of dominoes. This non-convex and non-smooth constrained problem is efficiently solved by a hybrid optimization framework, which integrates the alternating direction multiplier method (ADMM), aggregate function, modified real genetic algorithm (MGA), and non-uniform fast Fourier transform (NUFFT). Numerical simulations demonstrate that angular resolution varies with array geometry, even under the same aperture size. The proposed arrays outperform others with equal aperture size, exhibiting narrower MLW and lower Cramér-Rao bound (CRB), thereby enhancing angular resolution with fewer antennas and without preprocessing in standard single-snapshot 2D DOA estimation methods.