SS-CoSaMP-Based Channel Estimation for RIS-Assisted Multi-User Systems With Multi-Region User Distribution

Reconfigurable Intelligent Surface (RIS) is a pivotal enabling technology for sixth-generation (6G) wireless systems, yet the massive number of passive reflection elements in RIS introduces prohibitive pilot overhead during channel estimation—especially in multi-user scenarios where users are clustered in multiple angular regions. To address this challenge, this letter exploits the underutilized column structural sparsity induced by multi-region user distribution, a feature that prior works have not fully leveraged. First, a multi-region channel model is established to characterize the inherent correlation of user channels within the same angular region. Then, we propose a Structure-Sparse Compressive Sampling Matching Pursuit (SS-CoSaMP) algorithm, which decomposes cascaded channel estimation into two sequential stages: estimating the positions of sparse elements and recovering the channel matrix using the CoSaMP framework. Simulation results validate that the proposed SS-CoSaMP algorithm outperforms the exsiting methods in terms of Normalized Mean Squared Error (NMSE). Specifically, it achieves a 1–2 dB NMSE reduction compared to OMP and DS-OMP, and a 0.25 dB reduction compared to SS-OMP under the same signal-to-noise ratio (SNR), while significantly reducing pilot overhead.