Optimization of Loads for Antenna-Based Scattering Systems Using Feedforward Neural Networks

In this paper, we optimize the passive loads of a scattering system consisting of multiple coupled antennas using multi-layer feedforward neural networks. The developed in this study architectures, trained to solve a classification task, predict the load impedance values connected to unit antenna scatterers based on the bistatic radar cross-section of a structure. Trained networks exhibit potential in optimizing load values for scattering redirection among predefined directions. To demonstrate the applicability of the proposed method, two multi-layer feedfor-ward neural networks are trained and used to predict proper load impedances for different scattering objectives. Additionally, existing limitations of the method usage are discussed with the potential ways to mitigate them.