Downlink Non-Orthogonal Multiple Access (NOMA) in Poisson Networks

TL;DR

This paper analyzes downlink NOMA in Poisson networks, comparing ordering techniques, proposing resource allocation algorithms, and highlighting the impact of network parameters and clustering on performance and fairness.

Contribution

It introduces new models for UE clustering, compares ordering techniques, and develops algorithms for optimal resource allocation in NOMA Poisson networks.

Findings

Neither ordering technique is consistently superior.

Existence of an optimal number of UEs N for maximum sum rate.

Interference-aware clustering significantly improves performance.

Abstract

A network model is considered where Poisson distributed base stations transmit to $N$ power-domain non-orthogonal multiple access (NOMA) users (UEs) each {that employ successive interference cancellation (SIC) for decoding}. We propose three models for the clustering of NOMA UEs and consider two different ordering techniques for the NOMA UEs: mean signal power-based and instantaneous signal-to-intercell-interference-and-noise-ratio-based. For each technique, we present a signal-to-interference-and-noise ratio analysis for the coverage of the typical UE. We plot the rate region for the two-user case and show that neither ordering technique is consistently superior to the other. We propose two efficient algorithms for finding a feasible resource allocation that maximize the cell sum rate $\mathcal{R}_{\rm tot}$, for general $N$, constrained to: 1) a minimum throughput $\mathcal{T}$ for each UE, 2) identical throughput for all UEs. We show the existence of: 1) an optimum $N$ that maximizes the constrained $\mathcal{R}_{\rm tot}$ given a set of network parameters, 2) a critical SIC level necessary for NOMA to outperform orthogonal multiple access. The results highlight the importance in choosing the network parameters $N$, the constraints, and the ordering technique to balance the $\mathcal{R}_{\rm tot}$ and fairness requirements. We also show that interference-aware UE clustering can significantly improve performance.