Orthogonal Waveform Design for MIMO Radar Using Deep Learning

Abstract This paper explores the design of orthogonal waveforms for multiple-input multiple-output (MIMO) radar. We combine the peak sidelobe level (PSL) with the omnidirectional beampattern matching as the optimization objective. To address this challenging non-linear and non-convex problem, we propose an unsupervised optimization network based on residual network and attention mechanism. This approach harnesses the potent nonlinear fitting capabilities of neural networks to address the complex optimization problem. Simulation results indicate that the waveforms designed by the algorithm proposed in this paper exhibit better orthogonality than those designed by the traditional algorithm.