Expectation Propagation-Based Active User Detection and Channel Estimation for Massive Machine-Type Communications

Abstract

In massive machine-type communication (mMTC), by utilizing sporadic device activities, compressed sensing based multi-user detection (CS-MUD) can be used to recover sparse multi-user vectors in the grant-free uplink non-orthogonal multiple access (NOMA) environments. In CS-MUD, the channel state information (CSI) between each active device and the basestation should be estimated before the symbol detection. In this paper, we propose a novel Bayesian joint active user detection (AUD) and channel estimation (CE) method based on the expectation propagation (EP) algorithm. The proposed method finds the best Gaussian approximation for the computationally intractable posterior distribution of the sparse channel vector using iterative EP parameter update rules. Using the approximated distribution, identification and CSI estimation of active devices are jointly performed. We show from numerical simulations that the proposed technique greatly improves the performance of AUD and CE.