AI-Assisted Global Optimization for Solving Inverse Scattering Problems

Abstract

The solution of inverse scattering (IS) problems supported by artificial intelligence (AI) is addressed. An innovative solution strategy based on the System-by-Design (SbD) paradigm is proposed for the computationally-efficient exploitation of a global optimization strategy for minimizing the data mismatch cost function. Towards this end, a suitable selection, customization, and interconnection of SbD functional blocks is adopted. Moreover, the computationally-unaffordable repeated evaluation of each trial solution during the optimization is bypassed thanks to the exploitation of a digital twin (DT) based on the learning-by-examples (LBE) paradigm. An illustrative numerical example is shown to prove the effectiveness and computational efficiency of the proposed solution strategy when dealing with 2D free-space microwave imaging (MI) scenarios.