This paper investigates the application of nonorthogonal multiple access (NOMA) in millimeter-Wave (mmWave) communications to address resource allocation issues for multiuser downlink transmission. Inspired by the K-means clustering algorithm, a NOMA user grouping policy is first employed in accordance with the channel correlation between users. The first-stage decoding order for each NOMA group is then determined by merely utilizing the users’ channel gain to formulate a joint power allocation and hybrid beamforming optimization problem. Specifically, we use signal-to-leakage-plus-noise ratio (SLNR) as the performance index of the optimization problem to reduce the computational complexity, because it can decouple the design of the beamforming and power allocation issues so they can be executed iteratively. Under given beamforming matrix, the power allocation issue is expressed as a quadratic programming (QP) problem, and then transformed into a convex problem by introducing an auxiliary-positive real variable to solve through the Lagrange multiplier method. On the contrary, the beamforming optimization problem is non-convex and very difficult to solve due to the constant modulus constraints in the hybrid architecture. However, the optimum solution of the ideal full-digital beamforming can be acquired primarily by generalized eigenvalue decomposition (GED). In this condition, two hybrid beamforming algorithms are proposed, in which normalized phase matching and equal gain transmission are performed respectively in the analog domain. On this basis, a second-stage decoding order can be employed according to the effective channel gain of users to perform successive interference cancellation (SIC) successfully. Numerical results show that the proposed joint power allocation and hybrid beamforming algorithms are superior to the classical mmWave-NOMA and orthogonal multiple access (OMA) schemes not only in lower implementation complexity, but also in better spectrum efficiency and energy efficiency performance.

In this article, we study resource optimization issues for hybrid energy harvesting massive multiple-input–multiple-output (MIMO) systems, where renewable energy harvested from surrounding environments is introduced as additional energy supply to the base station (BS) together with the grid power. Our purpose is to maximize system energy efficiency with the preferential use of renewable energy under several practical restrictions, such as limited battery capacity, energy causality, quality-of-service (QoS) requirement, and transmit power constraints. An offline policy is first presented under ideal assumptions, i.e., noncausal knowledge of the channel state information and energy harvesting dynamics are known in advance, which causes it to become a performance upper bound. In particular, the offline policy is designed in an iterative manner with the use of the fractional programming technique. Moreover, inspired by the offline policy, we exploit statistical information to present two practical online algorithms. Specifically, the Markov chain method predicts prospective information through use of the finite-state Markov model, whereas the timing detection scheme focuses only on the resource optimization of the current time slot. Numerical results illustrate that both online algorithms perform closely to the offline policy, which indicates that they can achieve a good tradeoff between performance and implementation complexity.

In this paper, we consider a multi-reconfigurable intelligent surface(RIS)-assisted multi-cell communication system in which we jointly optimize the base station (BS)-RIS-user association and RIS deployment locations to maximize the ergodic capacity. In order to derive deterministic approximate expressions for the ergodic capacity, active and passive beamforming are first optimized to maximize the signal power of the relevant path. Then, we propose a two-level nested algorithm to address this mixed-integer nonlinear programming (MINLP) problem. Finally, the effectiveness of the proposed algorithm is validated through numerical simulations.

In recent years, deep learning (DL) has seen extensive applications in multiple-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) systems. In particular, DL has been integrated into receivers to enhance signal processing capabilities, exemplified by the DeepReceiver. However, such integrations typically train on static scenarios, resulting in suboptimal performance in dynamic real-world conditions and limiting broader applicability. To overcome these challenges, we introduce a DL-based receiver that employs transfer learning (TL) in MIMO-OFDM systems for rapid and accurate signal detection after sufficient training. Especially, we develop a novel binary classifier to accelerate training processes. Furthermore, we propose two TL strategies tailored for the DeepReceiver, designed to facilitate quick retraining under changing conditions. Simulation results confirm that our proposed receiver not only achieves superior bit error rate performance but also adapts more effectively to dynamic channels.

We investigate power optimization for throughput maximization in massive multiple-input multiple-output (MIMO) downlink networks with hybrid energy harvesting transmitter. In particular, a finite-capacity energy harvester is employed at the base station in conjunction with the fixed power grid for system energy supply. With availability of the non-causal knowledge about the channel state information (CSI) and the energy harvesting process, the throughput maximization is first formulated as a typical convex optimization problem, subject to a set of battery storage and power constraints. Specifically, this optimization is also featured by taking into account the circuit energy consumption that can not be neglected in massive MIMO systems with a large quality of radio frequency (RF) chains. Mathematically, Lagrange dual decomposition technique can be used to derive the optimal power variables. However, due to its offline calculation characteristics, the proposed approach is hard to implement in practice and can only serve as a performance upper bound. To facilitate real world application, we then heuristically present an online scheme that merely utilizes the causal information of the CSI and the statistics of the energy harvesting process, without significantly degrading the performance. Finally, simulation results are shown to verify the effectiveness of the proposed algorithms.

Utilizing the classical uniform arrays in massive multi-input multi-output (MIMO) communications will lead to rank-deficiency of channel matrix under line-of-sight (LOS) propagation scenarios, which can be further manifested as the uneven distribution of channel eigenvalues, resulting in the inability to support multi-stream data transmissions. In this paper, we are committed to addressing this problem by the topology design of non-uniform planar array (NUPA) under a generalized scene where the receiving and transmitting arrays are misaligned placed with certain placement angles and a relative rotation offset angle. Specifically, the geometric structure relationship of the arrays must be investigated to derive a general expression for the channel capacity with this NUPA deployment. By applying the singular value decomposition (SVD) method and analyzing the relationship between the eigenvalues and the misaligned angles, the impact of misaligned angles on channel capacity can be disclosed, i.e., channel capacity under specific transmission distance is related to the offset angle but irrelevant to the placement angles. The NUPA design guidelines can then be acquired by using the Kronecker product. Numerical results confirm our theoretical analysis and demonstrate the superiority of the proposed NUPA topology in promoting channel capacity.