SLL Minimization for Linear Uniformly Excited Sparse Arrays Using Gradient Descent

In this paper, we apply a gradient descent-based method to the synthesis of uniformly excited sparse phased arrays. The effectiveness of gradient-based optimization in this domain is highly dependent on the precision of the Peak Sidelobe Level (PSLL) estimation. Conventional grid-based calculation methods fail to provide the necessary accuracy, introducing numerical noise that prevents efficient convergence. To address this, we implement a grid-less framework for radiation pattern analysis. This approach achieves near-machine precision and enhances computational performance compared to traditional discrete grid search. Based on this continuous evaluation, we develop a modified gradient descent strategy capable of synthesizing large-scale arrays. The proposed method manages high-dimensional search spaces and outperforms heuristic algorithms in terms of both convergence speed and the level of sidelobe suppression.