1: \begin{abstract}

2: %	Extremely large antenna arrays (ELAA) are regarded as a promising technology for supporting sixth-generation (6G) networks. However, the large number of antennas significantly increases the computational complexity in precoding design, even for linearly regularized zero forcing (RZF) precoding. Moreover, the near-field spatial non-stationarity effect caused by ELAA should be considered in precoding design, as users primarily receive signal power from some subarrays of the entire large array, referred to as the visibility region (VR). In this paper, we investigate low-complexity iterative precoding design for ELAA-based multiuser systems by exploiting the properties of spatial non-stationarity. Inspired by the classical Kaczmarz-type algorithm and leveraging the features of VR information in near-field ELAA communication, we propose a VR-based orthogonal greedy random Kaczmarz (VR-OGRK) algorithm, which is proved to converge globally at an exponential rate. Next, by utilizing the information from multiple hyperplanes in each iteration, we extend the original Kaczmarz-type algorithm into the aggregation hyperplane Kaczmarz (AHK) algorithm, which further enhances the convergence rate. Finally, to fully exploit the VR information and reduce the computational complexity, we propose a  VR-based orthogonal aggregation hyperplane Kaczmarz (VR-OAHK) algorithm, and prove its  global-exponential convergence rate. Simulation results show that the two proposed algorithms achieve faster convergence and lower computational complexity than benchmark algorithms, and yield very similar rate performance to RZF.

3: %\end{abstract}

4: