Deep Unrolling-Based Joint Active User Detection and Channel Estimation in Distributed Antenna Systems

This paper addresses the challenge of joint active user detection (AUD) and channel estimation (CE) for grant-free random access within massive machine-type communications (mMTC) enabled distributed antenna systems (DAS). Sparse Bayesian learning (SBL) integrated with variational Bayesian inference (VBI) methods exploit channel sparsity effectively and are commonly employed for AUD and CE. Nevertheless, these techniques generally require matrix inversion operations and numerous iterations, leading to high computational complexity. To mitigate these issues, we propose a deep unrolling neural network based on a fast inversion algorithm, introducing learnable parameters to accelerate convergence. Additionally, we consider the pilot collision in grant-free random access, which is often overlooked in most literature, and presents a pilot collision handling method that leverages the spatial sparsity of distributed antennas. Simulation results validate the reliability of the proposed network and the method for pilot collision processing.