Document
Metadata
Authors
Mohamed Ahmed Adan, Mohamed Osman Omar, Mohamed Elmi Ahmed, Hafso Mohamed Mohamoud, Abdinasir Hirsi, Abdi Mohamed
Title
Prediction and Optimization of Mutual Coupling in Closely Spaced MIMO Arrays Using Deep Neural Networks
Journal
2026 International Conference on Artificial Intelligence for Sustainable Engineering and Innovation, AISEI 2026
Year of Publication
2026
Abstract
Mutual coupling is a major challenge in compact MIMO antenna arrays, where limited space forces antennas to be placed at sub-wavelength distances. This leads to degraded performance, including reduced radiation efficiency, lower channel capacity, and higher inter-port correlation. Conventional approaches such as analytical models, full-wave electromagnetic simulations, and physical decoupling structures are often computationally expensive and less effective for very compact designs. To address this, we propose a deep neural network (DNN)-based framework for fast prediction and optimization of mutual coupling. The model is trained using a large dataset of antenna geometries and corresponding S-parameters obtained from EM simulations. This approach enables rapid exploration of the antenna design space and reduces reliance on time-intensive simulations. An optimization strategy is also included to identify reliable design parameters while preserving key dimensions. The proposed method significantly reduces computational time and achieves high prediction accuracy for array gain (R2>0.99). However, it performs poorly in predicting mutual coupling (R2≈-10), suggesting that mutual coupling behavior is highly complex and not fully captured by the current feature set. © 2026 IEEE.
DOI
10.1109/AISEI68628.2026.11572879
Category
Engineering, Computing, and IT