Covariance Based Joint Activity and Data Detection for Massive Random Access with Massive MIMO

Phase Transition Analysis for Covariance Based Massive Random Access with Massive MIMO

“…Recently a number of non-orthogonal sequence based random access schemes for mMTC have been proposed, e.g., the two-phase grant-free random access [5], the grant-free random access with data embedding [13] or data spreading [19]. The non-orthogonal sequences can be used as signatures for active device detection, e.g., [8], [11], [20], as codewords for data transmission, e.g., [12], [21], [22], or as both, e.g., [1], [13]. By detecting which sequences are transmitted, the BS acquires the identification of the active devices, or/and the data bits.…”

“…Conventional cellular systems provide random access for human-type communications by employing a set of orthogonal sequences, from which every active device randomly and independently selects one sequence to transmit as a pilot for This work was supported by the Natural Sciences and Engineering Research Council (NSERC) of Canada. The materials in this paper have been presented in part at the IEEE International Conference on Communications (ICC), Shanghai, China, May 2019 [1], and at the Asilomar Conference on Signals, Systems, and Computers, Pacific Grove, CA, USA, November 2019 [2].…”

Hyperparameter-Free Receiver for Grant-Free NOMA Systems With MIMO-OFDM

“…PROPOSED METHOD In this section, we propose two hyperparameter-free receivers. Unlike the existing approaches, including that of [12], our proposals exploit the row-sparsity instead of the block sparsity of X in (5). After formulating the problem for the user activity and channel estimation as an maximum likelihood (ML) problem [3], we reveal a nontrivial mathematical connection between ML [3] and SPARROW [15], which is the ℓ 2,1 mixed-norm minimization problem.…”

“…, andX is the matrix whose each row obeys the complex standard Gaussian distribution. Based on the model given by (6), the ML estimation problem can be represented as [3], [5] (7) whereR = YY H / denotes the sample covariance matrix, and D + denotes the set of non-negative diagonal matrices.…”

“…However, they require updates of many messages at each iteration, resulting in prohibitive computational complexity at the BS as the number of system parameters increases. Besides, the conventional receivers for GF-NOMA systems, e.g., [3]- [5], require an exhaustive search of their design parameters called hyperparameters for all (time-varying) system parameters such as signal-to-noise ratio (SNR) and user activity ratio. Doubtlessly, it would enforce a huge overhead to obtain those system parameters and tune the corresponding hyperparameters before transmission in practice.…”