AI-Assisted Global Optimization for Solving Inverse Scattering Problems

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.