Efficient synthesis of concentric ring arrays with machine learning approach

Abstract

In this paper, we propose an efficient synthesis method based on two-stage optimization to synthesize the concentric ring array (CRA) through exploiting the potential benefits of machine learning approach, called K-Nearest Neighbor-Bagging (K-BAG). In the first stage, the training dataset of K-BAG is obtained through an optimization algorithm to train the K-BAG. After that, in the second stage, the randomly selected geometrical parameters of CRA are input to the trained K-BAG, then the structural parameters and correspondingly simulated results are obtained simultaneously in output, which can acquire better solutions compared with that of only considering optimization algorithm. The proposed method improves simulation efficiency and K-BAG robustness while guaranteeing the peak sidelobe level compared with other optimization algorithms of previous researches.