Unsourced Random Access for mmWave Massive MIMO-OFDM Over Doubly-Selective Channel

Time-frequency doubly-selective characteristics are often exhibited by wireless transmission channels, due to hardware imperfections and the relative mobility of devices. This paper investigates a massive unsourced random access (URA) scheme for the millimeter-wave (mmWave) band in the presence of a doubly-selective channel. By utilizing the sparse path and beam aggregation properties of mmWave communications, the equivalent channel can be modeled as a multi-dimensional block-sparse matrix. This modeling is achieved in the code-time-frequency (CTF)-angle domain through segment-based compressed coding. The joint channel estimation and segment data recovery is then achieved using a pattern-coupled sparse Bayesian learning (PCSBL) with generalized approximate message passing (GAMP) algorithm. A clustering method based on similar channel features is proposed to concatenate different data segments. Numerical simulation results validate the effectiveness of the proposed scheme, demonstrating a 6 dB increase in energy efficiency compared to tree-based coded compressed sensing (CCS) scheme.