Direct Detection of Object Parameters From Raw Data in MIMO-SAR Imaging at 235–270 GHz via Inverse Simulation

Synthetic imaging at millimeter wave (mm-wave) and terahertz (THz) frequencies with Multiple-Input Multiple-Output (MIMO) systems is a promising technique for non-invasive sensing in fields such as medical diagnostics, security screening, and materials characterization. Conventional synthetic imaging relies on reconstruction methods, such as back-projection, as a mandatory prerequisite for further scene analysis. However, these techniques often suffer from undersampling and reconstruction artifacts, limiting accuracy and interpretability. In this paper, we present a novel reconstruction-free approach for accurate, direct object parameter estimation in MIMO synthetic aperture radar (SAR) systems, based on the concept of inverse simulation. Our method uses a differentiable simulation of electromagnetic wave propagation to directly estimate object properties such as object location, geometric parameters, and material model parameters from raw measurement data. A tailored objective function is introduced to align simulated sensor responses with actual measurements, enabling the end-to-end optimization of scene parameters via inverse simulation. By completely bypassing traditional image reconstruction, the proposed framework offers robust and interpretable performance, even in reflective environments. We validate our method experimentally using a 10 × 31 MIMO SAR system operating at 235–270 GHz. Experiments illustrate the capability to localize spheres with sub-100$\mathrm{\mu }$m precision, demonstrating an order of magnitude improvement over standard back-propagation analysis (BPA), despite significant clutter in the image data. These results highlight the potential of inverse simulation for next-generation mm-wave and THz sensing instrumentation and computational imaging.