Adaptive User Pairing in Multi-IRS-aided Massive MIMO-NOMA Networks: Spectral Efficiency Maximization and Deep Learning Design

• Published in: IEEE transactions on communications Vol. 71; no. 7; p. 1
• Main Authors: Perdana, Ridho Hendra Yoga, Nguyen, Toan-Van, An, Beongku
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.07.2023; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Complexity theory; Deep learning; Deep learning framework; Integer programming; intelligent reflecting surface (IRS); Iterative algorithms; Iterative methods; Massive MIMO; massive MIMO networks; Maximization; Maximum power; MIMO communication; Mixed integer; NOMA; non-convex optimization; non-orthogonal multiple access (NOMA); Nonorthogonal multiple access; Optimization; Phase shift; Quality of service; Resource management; Resource utilization; Signal quality; Spectral efficiency; user pairing
• ISSN: 0090-6778; 1558-0857
• DOI: 10.1109/TCOMM.2023.3277533

In this paper, we propose an adaptive user pairing (AUP) scheme in multi-intelligent reflecting surface (IRS)-aided massive multiple-input multiple-output (MIMO)-non-orthogonal multiple access (NOMA) networks. In the AUP scheme, two users with different channel conditions are selected for user pairing while multiple IRSs assist to improve received signal quality at users. We consider the problem of jointly optimizing the precoding matrix, the phase shift of IRSs, and the user pairing element to maximize the overall spectral efficiency (SE) subject to the maximum power budget at the base station (BS) and user-specific quality-of-service (QoS). The SE problem formulated as the maximization of non-concave functions involves a mixed-integer program, which is very challenging to solve optimally. To tackle this problem, we first relax the user pairing elements to be continuous and then transform the formulated problem into an equivalent non-convex problem with a more tractable form. We then apply the iterative algorithm (IA) with low complexity to guarantee convergence at a relative optimum. Towards real-time optimization, we propose a deep learning (DL) framework to predict the optimal solution of the precoding matrix, the phase shift of IRSs, and user pairing elements according to the user's locations and channel gains. Compared to the conventional optimization method, the DL-based optimization framework can achieve the optimal solution within a very short time via an efficient inference process. Numerical results verify that the proposed algorithm improves the SE over state-of-the-art approaches. Moreover, the effects of essential parameters such as the total BS transmit power, the number of UEs, IRSs, and BS's antennas on the system are discussed and evaluated to show the effectiveness of the proposed scheme in balancing resource utilization.