Unsourced Massive Random Access Scheme Exploiting Reed-Muller Sequences

Abstract

The challenge in massive Machine Type Communication (mMTC) is to support reliable and instant access for an enormous number of machine-type devices (MTDs). In some particular applications of mMTC, the access point (AP) only has to know the messages received, but not where they source from, thus giving rise to the concept of unsourced random access (URA). In this paper, we propose a novel URA scheme exploiting the elegant properties of Reed-Muller (RM) sequences. Specifically, after dividing the message of an active user into several information chunks, RM sequences are used to carry those chunks, for exploiting the vast sequence space to improve the spectral efficiency, and their nested structure to enable reliable and efficient sequence detection. Next, we further explore a novel structural property of RM sequences for designing sparse patterns which carry part of the information and serve as the hints of coupling the information chunks of a single user. The factors affecting the performance of our slot-based RM detection are characterized. Besides, the complexity of the proposed message stitching method is analyzed and compared to the commonly used tree coding approach. Our simulation results verify the enhanced performance of the proposed URA scheme in error probability and computational complexity compared to the existing counterpart.

An Efficient Two-Stage SPARC Decoder for Massive MIMO Unsourced Random Access

“…Various schemes has been proposed for URA in AWGN channels, such as T -fold ALOHA [7], coded compressive sensing [8,9] and sparse regression code (SPARC) [3]. Considering the channel fading in wireless system, the T -fold ALOHA scheme [10,11] and the Reed-Muller based scheme [12] were proposed for URA 1 Here, as in [2][3][4], we assume that users are synchronized.…”

“…Let µ be given in (20), η be given in (7), and the sequence {γ t } t≥0 be defined by IHT in (12). For some constant step size α, if {a k } N k=1 satisfy the MIP condition…”

“…The second stage of Algorithm 1 (i.e., ML or NNLS) has been analyzed in the previous work [4,18], thus we focus on the first stage of the algorithm. By (19), we know that selecting K largest components in {Y k } N k=1 is equivalent to the one-step IHT in (12). Consider that D, M are reasonably large and K = K. The following theorem shows that the estimation of one-step IHT satisfies supp (γ 1 ) = S with overwhelming probability.…”

“…3 demonstrates the related computational complexity in terms of processing time in second. IHT in (12), ML…”

“…Suppose that the 3s-th restricted isometry constant of the matrix A D ∈ C D 2 ×N satisfies δ 3s ≤ 1 √ 3 . Then, for any s-sparse γ and δ in (10) with δ 2 ≤ η, the sequence {γ t } t≥0 defined by IHT in (12) with some constant step size α satisfies Therefore, by (34) we have…”