MIMO Radar Polyphase Waveform Design via Optimal Loss Function-Based Soft Quantization

Polyphase waveform design for the minimization of Integrated Sidelobe Level Ratio (ISLR) is the key technology in Multiple-Input Multiple-Output (MIMO) radar systems. Due to the discrete phase constraint, the problem is non-convex and challenging to solve. Existing methods often rely on experience-based hard quantization with fixed thresholds, leading to limited performance due to the hard quantization. To address this issue, we propose the Optimal Loss Function-Based Soft Quantization (OLF-SQ) method with adjustable quantization thresholds. Firstly, the Complex Circle Manifold (CCM) space is constructed to satisfy the constant modulus constraint, and then the Gradient Descent (GD) model-driven network layer based on the CCM is devised to obtain the continuous waveform. Secondly, the soft quantization function is derived with adjustable quantization thresholds, and then the soft quantization network layer is devised to obtain the discrete waveform, where the quantization thresholds are learned by the unsupervised learning network. Compared with the existing methods, the proposed method has better performance in terms of ISLR and beampattern shaping.