Efficient Near-Field Beam Focusing Merging Orthogonal Matching Pursuit and CVX for Large Intelligent Surface Applications

Abstract

In this paper, an efficient near-field beamforming method is proposed to support the large intelligent surfaces (LIS) that are expected to be widely deployed in 6G networks. This approach avoids directly applying convex (CVX) optimization for sparse selection in large-size array matrices, as such methods often lead to excessive computational time due to blind searching to satisfy a series of objective functions. First, based on the objective function, we prioritize a key component and employ the orthogonal matching pursuit (OMP) method to pre-select potential sparse target positions. To ensure focal symmetry, a coordinate mirror symmetry approach is adopted, meaning that selection is performed only in the first quadrant, while the remaining quadrants are determined through mirror symmetry relative to the first quadrant. This significantly reduces computational complexity at an early stage. Next, CVX is applied based on the pre-selected sparse array. Once a predefined threshold is met, a solution is obtained that satisfies the constraints of the beamfocusing. The results demonstrate that, compared with conventional methods, this approach improves efficiency by 15.12 times with 121 elements and 96.73 times with 441 elements. The proposed method demonstrates not only satisfactory performance but also considerable potential as a beam focusing technique for large-scale near-field array systems.