Displaced Subarray for Uniform Partitioning and Sparse Feeding

Large-scale array beamforming has demonstrated enhanced communication performance. However, as the operating frequency increases, the full array has to be divided to reduce the hardware cost. In this paper, we consider applying the displaced subarray to mitigate the grating lobe issues caused by uniform subarray partitioning and sparse feeding. Specifically, we calculate the gradient of the grating lobes with respect to the subarray position. A hybrid optimization framework integrating momentum-accelerated gradient descent with cosine annealing learning rate adaptation is developed, where systematic displacement parameter optimization is performed under dual constraints of array boundaries and non-overlapping subarray configurations. Numerical results demonstrate that the proposed algorithm significantly reduces the grating lobes compared to other benchmark schemes.