Optimization and Machine Learning for Antenna Array Healing

Maximizing antenna array performance has been an area of active research for decades. To this end, numerous statistical- and optimization-based approaches have demonstrated significant improvements to array bandwidth, scanning range, and side lobe levels, among other performance targets compared to conventional array design techniques. However, antenna array elements can fail due to mechanical degradation, environmental factors, electrical interference, or through unintended destructive means. Furthermore, element failures can lead to significant performance degradation and even the creation of spurious beams radiating in unwanted or unsafe directions. Therefore, the ability to overcome array failures is of paramount importance. While existing approaches based on statistical and global optimization methods have demonstrated success in array healing, deep learning techniques offer the potential for near-instantaneous performance recovery. The paper introduces our deep-learning accelerated framework for arbitrary on-demand array-healing.