Electromagnetic machine learning for estimation and mitigation of mutual coupling in strongly coupled arrays

With the recent advancement in baseband integrated circuits and radio frequency (RF) technology, there is a strong demand to minimize the size of array-based microwave systems. However, array aperture miniaturization often leads to strong mutual coupling in the arrays. To date, mitigating and/or compensating the mutual coupling (nonlinear) effect remains a technical challenge. For instance, realizing microwave orientations in the presence of strong mutual coupling is a challenge. In this paper, we propose a unified electromagnetic machine learning (EMML) technique for overcoming this challenge. Antenna current's Green function (ACGF) is employed for characterization of the EM response of elements in form of digital signal processing (DSP) module with the inclusion of mutual coupling effect among the elements. An EMML framework is then formulated and combined with the DSP module based on ACGF, to consequently come up with a nonlinear mutual coupling mitigation framework that provides better decoupling capability than the previous linear techniques. EMML is verified by application to a multiple signal classification (MUSIC)-based direction-of-arrival estimation scenario, and compared with the conventional methods. The results obtained show the validity, effectiveness of the EMML method. This method can be further used in coupled array-based beam forming, steering, and nulling applications.