Uncoupled Unsourced Random Access for Near-Field Spatial Non-Stationary XL-MIMO

In this correspondence, we propose a uncoupled compressed sensing scheme in near-field spatial non-stationary extremely large-scale multiple-input multiple-output unsourced random access communications. Concretely, first, for multiple user detection in each slot, a codeword-collision-aware two-stage orthogonal matching pursuit algorithm is proposed, which can utilize the three-level sparsity of the equivalent channel coefficient, and is able to resolve codeword collision adaptively via estimation of multiple visibility regions (VRs) and the angles of arrivals (AoAs) for collided codeword. Then, the$K$-medoids-based clustering is employed to achieve efficient message stitching by utilizing the invariant characteristics of VR and AoA during one transmit frame consisting of multiple slots. Finally, numerical results verify the effectiveness of the proposed method.