Hardware Validation of Unimodular Waveforms Using an Original Autoencoder Designed for Code Division Multiplexing MIMO Radar Systems

Multiple input multiple output (MIMO) radar is a technology for realizing next-generation radar systems. Realizing MIMO radar by dividing the aperture of a phased array antenna can search multiple directions simultaneously, but preventing interference between apertures, the waveforms must have low cross-correlation. Furthermore, to apply for radar systems, the alphabet size needs to be finite owing to hardware constraints. To design waveforms sufficient for these requirements, we proposed a method of designing sequences of finite alphabet sizes with good cross-correlation properties using the residual network (ResNet) autoencoder. Applying the proposed method, waveforms with cross-correlation levels superior to those of cyclic-algorithm new (CAN) and Hadamard sequence with a pseudo-noise (Hadamard+PN) can be designed; 128-phase shift keying (PSK) and 256-PSK achieved lower than -50 dB. Since, despite this background, the applicability of the proposed method to actual machines had not been verified, in the present work, to confirm the effectiveness of this method, we performed numerical simulation and actual machine verification using waveforms designated by the proposed method.