Joint Activity Detection and Channel Estimation in MIMO Grant-Free Random Access Networks

Massive machine communication is predicted to provide widespread and unparalleled connectivity for cellular Internet of Things (IoT) applications via multiple-input multiple-output (MIMO) and grant-free random access (GF-RA) techniques. Compressed sensing (CS) has been widely advocated to support massive connectivity due to the bursty nature of traffic transmission. In this paper, the joint activity detection and channel estimation in MIMO-enabled GF-RA system is formulated as a block single measurement vector (SMV) problem and efficiently addressed by using Bayesian-based CS algorithms. First, the pattern coupled sparse Bayesian learning (PCSBL) and block sparse Bayesian learning (BSBL) algorithms are introduced to solve this problem, where the potential block sparsity properties induced by multi-antenna reception are exploited by assigning the structured hyperpriors. Then, by embedding the Generalized Approximate Message Passing (GAMP) technique into the PCSBL and BSBL frameworks to enable effective approximation of posterior distributions, we propose two computationally efficient Bayesian learning algorithms, i.e., GAMP-PCSBL and GAMP-BSBL. Fortunately, the proposed Bayesian algorithms allow automatic learning of block sparse solutions without requiring noise level and user sparsity ratio as explicit conditions. Simulation results show that the proposed algorithms provide improved performance gains over the standard CS-based methods.