Orthogonal Waveform Design for MIMO Radar Based by Deep Learning

In MIMO (Multiple-Input Multiple-Output) radar systems, optimizing phase-coded waveform sequences with low autocorrelation sidelobes and low inter-correlation peaks is crucial for improving the detection performance of the system. Traditional waveform design methods often fail to effectively capture the complex dependencies within the signal sequence, making it difficult to achieve globally optimal sidelobe suppression. To address this issue, this paper proposes an optimization method for orthogonal phase-coded signal sets based on the Long Short-Term Memory (LSTM) model. The method can model long-range dependencies in the waveform sequence and effectively capture the complex patterns in the phase-coded sequences. The model learns to minimize the loss function of autocorrelation and inter-correlation sidelobes, resulting in waveforms with good sidelobe suppression in the time domain. Compared with traditional optimization methods, the LSTM-based waveform design method can adaptively adjust the phase sequences, effectively reducing autocorrelation and inter-correlation sidelobes while maintaining good signal orthogonality. Experimental results verify that the generated waveforms significantly reduce sidelobe levels and enhance the system’s signal processing performance, with broad application prospects.