Design and optimization of a 77 GHz absorber for autonomous vehicles using artificial neural networks

Abstract

This work presents the design and optimization of a Frequency Selective Absorber (FSA) operating at 77 GHz, aimed at automotive radar and autonomous vehicle systems applica tions. The methodology combines computational modeling techniques, such as the Finite Element Method (FEM) and the Finite Integration Technique (FIT), with artificial neural networks (ANNs), specifically a multilayer perceptron (MLP), to optimize the geometric and material parameters of the absorber. The FSA structure comprises a Rogers Duroid RT5880 substrate and a copper metallic ground plane. The modeling employs an equivalent circuit with resonant elements (inductors, capacitors, and resistors) to describe the filter’s behavior. We trained the ANN with simulation data, achieving high accuracy in predicting the optimal parameters, such as periodicity, substrate thickness, and aperture dimensions. Results demonstrate high absorber efficiency, with an absorption rate of 99.99% at the central resonance frequency of 77 GHz. The comparison between the ANN predictions and numerical methods (HFSS and CST Studio Suite) shows excellent agreement, validating the robustness of the proposed approach. The study concludes by highlighting the practical impact of the optimized design on enhancing the performance of automotive systems, emphasizing the potential of integrating ANNs into high-frequency electromagnetic device design.