Joint Power Control and User Association for NOMA-Based Full-Duplex Systems

TL;DR

This paper proposes joint power control and user association algorithms for NOMA-based full-duplex systems to enhance spectral efficiency and fairness, introducing tensor models and low-complexity solutions with near-optimal performance.

Contribution

It introduces a novel tensor-based optimization framework for joint user association and power control in NOMA-FD systems, along with low-complexity algorithms and a benchmark optimal solution.

Findings

Proposed algorithms outperform conventional FD and half-duplex schemes.

Algorithms achieve data rates close to brute-force optimal solutions.

Tensor model effectively optimizes UL decoding order and DL clustering.

Abstract

This paper investigates the coexistence of non-orthogonal multiple access (NOMA) and full-duplex (FD) to improve both spectral efficiency (SE) and user fairness. In such a scenario, NOMA based on the successive interference cancellation technique is simultaneously applied to both uplink (UL) and downlink (DL) transmissions in an FD system. We consider the problem of jointly optimizing user association (UA) and power control to maximize the overall SE, subject to user-specific quality-of-service and total transmit power constraints. To be spectrally-efficient, we introduce the tensor model to optimize UL users' decoding order and DL users' clustering, which results in a mixed-integer non-convex problem. For practically appealing applications, we first relax the binary variables and then propose two low-complexity designs. In the first design, the continuous relaxation problem is solved using the inner convex approximation framework. Next, we additionally introduce the penalty method to further accelerate the performance of the former design. For a benchmark, we develop an optimal solution based on brute-force search (BFS) over all possible cases of UAs. It is demonstrated in numerical results that the proposed algorithms outperform the conventional FD-based schemes and its half-duplex counterpart, as well as yield data rates close to those obtained by BFS-based algorithm.