MIMO Radar Waveform Design via Deep Learning

Abstract

Constant modulus (CM) waveform design with the high signal-to-interference and noise ratio (SINR) has been recognized as a key issue in Multiple-Input Multiple-Output (MIMO) radar system. In this paper, we propose a novel approach based on deep learning (DL) for waveform design with high SINR under CM and low integrated sidelobe levels (ISL) constraints. The resulting design is a nonconvex optimization problem and can not be solved directly. To solve this problem numerically, we divide it into two subproblems. We solve one subproblem by the DL method to obtain the CM waveform with low ISL. For the other subproblem, we relax it into a semi-definite programming (SDP) problem, and a covariance matrix is obtained by solving this SDP problem. Finally, the waveform matrix and the waveform covariance matrix are combined in an interesting way to obtain the desired waveform. Numerical results show that our proposed algorithm has better performance compared with the recent introduced methods.