Two-Dimensional Off-the-Grid Direction of Arrival Estimation for Millimeter-Wave Imaging Radar

As an all-weather and all-day sensing tool, millimeter wave radar with compact size and small cost is gaining increasing attention in automotive intelligence. In particular, four-dimensional (4-D) imaging radar with range, doppler, azimuth and elevation resolving ability is playing an important role among automotive sensors. The sparse array adopted by 4-D radar and the requirements of high frame rate and high-resolution imaging put forward higher requirements for 2-D direction of arrival (DOA) estimation algorithm. In this paper, an off-the-grid compressed sensing (CS) algorithm for 2-D DOA estimation with sparse array is proposed. In the scheme, the sparse echo matrix together with the special spectrum in azimuth and elevation are estimated by solving the atomic norm minimization problem. Then, the respective DOAs of azimuth and elevation are obtained by Vandermonde decomposition. Finally, the estimated results are paired. The proposed algorithm can realize 2-D DOA estimation of sparse array using only a single snapshot, and the size of problem is significantly reduced compared with the gridding-based CS algorithm. The effectiveness of the proposed algorithm is verified by experimental analyses.