TRANM: Decoherenced DoA estimation for automotive radar using generalized sparse arrays

Abstract

This paper tackles the challenge of coherent single-snapshot direction-of-arrival estimation in automotive linear frequency modulated continuous wave (LFMCW) radar using a generalized sparse array. By leveraging atomic-norm minimization (ANM)-based interpolation and Toeplitz rearrangement, a TRANM framework is proposed to address the rank-deficiency issue in the range-Doppler domain. To further enhance computational efficiency, we reformulate the TRANM problem into an equivalent optimization with reduced dimensionality. The problem is then solved using the alternating direction method of multipliers, which provides an optimal solution via an iterative process. Numerical simulations validate that the proposed approach can accurately resolve coherent signals with improved degrees of freedom and achieve super-resolution, all while maintaining a low computational cost.