Non-orthogonal Multiple Access (NOMA) Techniques for Wireless Networks

無線網絡中的非正交接入技術

Student thesis: Doctoral Thesis

View graph of relations

Author(s)

Related Research Unit(s)

Detail(s)

Awarding Institution
Supervisors/Advisors
Award date30 Oct 2018

Abstract

Non-orthogonal multiple access (NOMA), usually involving power control (PC) and successive interference cancellation (SIC), has been widely discussed for fifth-generation (5G) cellular systems. However, we show in the introduction that the possible benefits of NOMA cannot justify its higher receiver cost under the considerations of idealized system control and accurate channel state information (CSI) estimation. The real advantage of NOMA lies in the systems without these idealized assumptions. In this thesis, we discuss NOMA in several non-idealized situations.

In Chapter 2, we consider an interleave-division multiple access (IDMA) scheme that provides an efficient NOMA solution under the considerations of non-ideal channel coding and non-ideal CSI estimates. IDMA with user-specific interleaving can facilitate iterative multi-user detection (MUD). We provide several IDMA implementation techniques, including system optimization schemes based on an evolution analysis, a low-cost shifting scheme to realize user-specific interleaving, detection techniques based on Gaussian approximation for MIMO systems and a combination of IDMA and data-aided channel estimation to refine CSI acquisition. We show that these techniques can offer excellent performance in different IDMA systems. We also briefly compare several 5G NOMA proposals to show that they share with IDMA a principle of reducing short cycles in their graphic representations to facilitate iterative detection.

In Chapter 3, we consider decentralized NOMA for random access (RA) situations. We first review a SIC-based randomized power control (RPC) technique, in which collisions of two RA users, referred to as type-2 collisions, can be resolved with a high probability. We then improve RPC in a small cell system, in which there are commonly two types of users, i.e. local users (LUs), who stay in the small cell with relatively low mobility, and visiting users (VUs), who enter and leave the small cell randomly with relatively high mobility. Priority is given to LUs, while VUs can transmit in parallel with LUs in a decentralized manner as long as the resultant outage is within a preset quality of service limit. Such a scheme is referred to as a semi-random access (SRA) scheme, in which an optimization problem is formulated to maximize the VU throughput. We also discuss an optimal-MUD-based generalized random access (GRA) scheme for RA users. We prove that the optimal strategy for the type-2 collision case has a simple discrete structure, where only four received power values should be randomly selected by each user with a certain probability. Relying on the insights from such a structure, we extend the approach to the general collision cases by a threshold-based binary PC scheme. We show that SRA and GRA can outperform existing alternatives.

In Chapter 4, we consider NOMA in a decode-and-forward two-path relay system without a direct source-to-destination path. In such a system, NOMA is used to manage the inter-relay interference at the relay receivers. We analyze the feasible power regions for two MUD strategies and then formulate a joint power and rate allocation problem to optimize the system. Numerical results show the advantages of the proposed NOMA schemes in power consumption and outage performance.

Finally, in Chapter 5, we conclude the thesis and outline possible future research directions.