Learned Complex Circle Manifold Network for MIMO Radar Waveform Design

• Published in: 2023 IEEE Radar Conference (RadarConf23) pp. 1 - 5
• Main Authors: Zhong, Kai, Hu, Jinfeng, Yuan, Ye, Zhu, Gangyong, Yu, Xianxiang, Cui, Guolong
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.05.2023
• Subjects: Adaptation models; Computational modeling; Constant modulus; Deep learning; Interference; LCCM-Net method; Manifolds; MIMO radar; Radar; Waveform design
• DOI: 10.1109/RadarConf2351548.2023.10149634

The waveform design for transmit beampattern with constant modulus constraint (CMC) is the key technology in Multiple Input Multiple Output (MIMO) radar systems. Usually, the existing methods mainly include two categories: classical statistical modeling methods, most of which relax the problem with performance degradation; and conventional data-driven Deep Neural Networks (DNNs) without considering the characteristic of CMC. We notice that the CMC can be geometrically interpreted as restricting the solution to a Complex Circle Manifold (CCM), then the model-driven Learned CCM Network (LCCM-Net) is proposed, where the gradient descent algorithm is unfolded as the network layer. Different from the existing methods, the proposed method considers both the characteristic of CMC without relaxation and adaptively adjusting the stepsize by the network. Compared with the existing methods, the proposed method can achieve SINR gain enhancement with less computational cost.