PAST: An Efficient Localization Algorithm for Atomic Norm Minimization in 3D MIMO Radar

The challenge of 3D localization has been a subject of investigation spanning decades, serving as a pivotal technology within expansive massive multiple-input multiple-output (MIMO) radar arrays. In this paper, we proposed an efficient localization method named parafac decomposition and atomic norm iteration shrinkage-thresholding (PAST) based on accelerated proximal gradient (APG) for 3D bistatic MIMO radar. By exploiting the 3D model of the received signal, the multi-parameter estimation problem is decoupled by tensor decomposition and converted into angle and velocity estimation problem respectively. Then, the atomic norm iteration shrinkage-thresholding is given to estimate the parameter for each dimension efficiently. By employing the APG, our approach enforces the ANM principle, eliminating the computationally intensive semidefinite programming (SDP). Simulation results show that the proposed PAST outperforms the traditional methods without parafac decomposition in the bistatic MIMO radar scenario. Moreover, our proposed method reduces computational complexity while achieving similar performance to SDP-ANM using tensor decomposition.