Massive MIMO-Enabled Semi-Blind Detection for Grant-Free Massive Connectivity

This paper studies the reliable support for massive machine-type communications and proposes an efficient semi-blind detection scheme for grant-free massive connectivity. In the proposed scheme, each active device directly transmits a very short reference signal along with its payload data in the uplink, without any access scheduling in advance. At the base station (BS), we develop a successive interference cancellation (SIC)-based semi-blind detection algorithm to detect active devices and their payload data. Specifically, benefitting from the large spatial dimensionality of the BS antenna array, the bilinear generalized approximate message passing algorithm is employed for joint channel and signal estimation (JCSE). In particular, we introduce an a priori refining strategy to leverage the structured sparsity of the massive access channel matrix for improved JCSE performance. Moreover, the inserted reference signal is utilized to resolve the inherent phase and permutation ambiguities. Besides, the idea of SIC is adopted for further improved detection accuracy, where the cyclic redundancy check and soft pilot-based channel refining are incorporated to prevent error propagation. Numerical results demonstrate that the proposed semi-blind detection scheme outperforms the state-of-the-art training-based coherent detection scheme when the same number of physical resources are occupied.