Knowledge-Guided Active-Base-Element Modeling in Machine-Learning-Assisted Antenna-Array Design

Abstract

Machine-learning (ML)-assisted antenna-array design methods suffer from a heavy computational burden. In this article, a knowledge-guided active-base-element-modeling (KG-ABEM) method is proposed for practical antenna array design to largely alleviate this burden. With the introduction of prior physical knowledge and trustworthy assumptions, the ML is only introduced to scenes where analytical methods cannot build accurate models, achieving great efficiency and accuracy improvement simultaneously compared with conventional ML-assisted modeling methods. First, to solve the intricate antenna array design tasks under mutual coupling and platform effects, the array is decomposed into an assembly of active base elements (ABEs) with different characteristics. Then, the active element patterns (AEPs) of these ABEs are approximately characterized by excitations of planar virtual subarrays. Finally, the ML method is introduced within the modeling process between the ABE geometry and corresponding excitations of the subarrays. By integrating the proposed KG-ABEM into an ML-assisted optimization (MLAO) algorithm, a fast and reliable array design is achieved. A set of design examples, including both linear and planar array design problems, are provided to validate the effectiveness and advantages of the proposed modeling and optimization method.