Low complexity Detector for massive uplink random access with NOMA in IoT LPWA networks

Abstract

We focus on the random uplink transmissions of an unknown subset of nodes, disseminated in a cell. Under the constraints of massive Machine Type Communication (MTC) in cellular Low Power Wide Area Networks (LPWAN) and Ultra Reliable Low Latency Communications (URLLC), we assume a low coordination with the receiver and the usage of Gaussian coded Non Orthogonal Multiple Access (NOMA). We then target direct data transmission and thus propose a low complexity optimal-based detection of the active users: the It-MAP. This algorithm relies on the Maximum A Posteriori (MAP) detector and, similarly to Orthogonal Matching Pursuit (OMP)like algorithms, proceeds by iteration to decrease its intrinsic complexity. We also show the gain of employing It-MAP rather than an OMP-based detection and the advantage of exploiting the possibility to tune the algorithm, in order to avoid either Missed Detection or False Alarm, depending on the wished trade-off between the reliability, the latency and the resource usage of the full transmission.

Coded random access for massive MTC under statistical channel knowledge

“…Moreover, the "iterative" (It) ML detector is an approximated ML algorithm as the detection is realised by accumulation of the detected users (one at each iteration), considering the detected users of the previous iterations as given for the current iteration. Its structure is the same as the proposed algorithm in [9]. If used together with the "targeted search" option, the iterations number thus matches the number of active nodes.…”

“…If used together with the "targeted search" option, the iterations number thus matches the number of active nodes. To provide a reference, the CER of the ML with iterative targeted search and channel knowledge (labelled "full CSIR" in the figures) [9] is also given as the detection is performed with a full CSIR assumption rather than its restriction to statistical knowledge.…”

“…Each node n ∈ Φ may transmit its own code c n (allocated in advance or randomly selected according to a specific procedure on which the BS and nodes agreed in advance). Following [8,9], the expression of the signal received at the BS is :…”

“…In the following we propose to compare the performance of a receiver exploiting either the random Gaussian codes as in [8,9] or the Grassmannian codes recently proposed in [11].…”

“…Indeed, any communication prior to the data or access request transmission would introduce high signalling overhead and would be irrelevant in the case of access request, since users willing to access the network would need to be already identified for allowing their channel estimation before they even send their access requests. For instance, in [7,8,9] the channel is considered as a known parameter at the receiver. Having a method able to work under statistical Channel State Information (CSI) only, is therefore a harsh constraint we deal with in this paper.…”