Gradient-Based PSLL Minimization for Uniformly Excited Sparse Linear Arrays

In this paper, we apply a gradient descent method to the synthesis of uniformly excited sparse phased arrays. Because the effectiveness of gradient optimization depends heavily on precise Peak Sidelobe Level estimation, we implement a grid-less framework for radiation pattern analysis to eliminate numerical noise. This approach achieves an absolute normalized intensity error of $10^{-16}$ while enhancing computational performance. Based on this continuous evaluation, the proposed approach effectively manages high-dimensional search spaces, demonstrating competitive performance against representative heuristic algorithms in sidelobe suppression for large-scale arrays.