A Machine Learning-Enabled Radiation-Scattering Integrated Design Approach for Low-Scattering Phased Arrays

To facilitate a rapid and synchronous design of radiation and scattering characteristics in low-scattering phased arrays, a machine learning (ML)-enabled radiation-scattering integrated design (MLE-RSID) approach is proposed in this letter. In this MLE-RSID approach, an inverse ML model is developed, wherein the radiation and scattering characteristics (| S ₁₁ | and reflection phase difference) of each two combined array elements are set as inputs, and their geometric parameters as outputs. Utilizing the proposed approach, designers can efficiently achieve phased arrays with on-demand radiation and scattering performances in near real-time. To validate the proposed approach, two antenna elements featuring a wideband scattering characteristic of (180° ± 37°) reflection phase difference and similar radiation characteristics are designed using the MLE-RSID approach, to construct a low-scattering 1 × 10 phased array.