Cross Validation Aided AMP-MMV Detector for Uplink Grant-Free NOMA

Grant-free non-orthogonal multiple access (NOMA) has been regarded as one of the key techniques for the next generation of wireless communications, especially suitable for massive machine-type communication (mMTC) scenarios. In addition, compressed sensing (CS) algorithms has been identified as a potential candidate for multiuser detection due to the natural sparsity of user activity in mMTC. The approximate message passing multiple measurement vector (AMP-MMV) algorithm with minimum mean square error (MMSE) denoiser is an efficient CS-based multiuser detection method, but it relies on prior knowledge of active user sparsity, which is difficult to obtain in practice. In this paper, we propose a cross validation (CV) aided MMSE-AMP-MMV (CVA-AMP-MMV) algorithm that can adaptively estimate the true user sparsity in frame- based uplink grant-free NOMA system. Active user detection is performed in parallel during the sparse estimation process without the need for additional detection steps. Simulation results show that the proposed CVA-AMP-MMV algorithm achieves the same performance as the MMSE-AMP-MMV algorithm at a slight complexity cost.