Millimeter wave (mmWave) system tends to have a large number of antenna elements to compensate for the high channel path loss. The immense number of BS antennas incurs high system costs, power, and interconnect bandwidth. To circumvent these obstacles, two-step hybrid precoding algorithms that enable the use of fewer RF chains have been proposed. However, the precoding schemes already in place are either too complex or not performing well enough. In this study, an equivalent channel hybrid precoding was proposed. The part from the transmitter RF chain to the receiver RF chain is regarded as equivalent channel. By reducing the dimension of channel matrix to the level of RF link number, base-band pre-coder is simply calculated from decomposing the equivalent channel matrix Hequ, which greatly reduces the complexity. Based on this novel precoding approach and convolutional neural network (CNN), a novel combiner neural network architecture was also proposed, which can be trained to learn how to optimize the combiner for maximizing the spectral efficiency with hardware limitation and imperfect CSI. Simulation results show that the proposed approaches achieve significant performance improvement.

In space-air-ground integrated networks (SAGIN), receivers experience diverse interference from both the satellite and terrestrial transmitters. The heterogeneous structure of SAGIN poses challenges for traditional interference management (IM) schemes to effectively mitigate interference. To address this, a novel UAV-RIS-aided IM scheme is proposed for SAGIN, where different types of channel state information (CSI) including no CSI, instantaneous CSI, and delayed CSI, are considered. According to the types of CSI, interference alignment, beamforming, and space-time precoding are designed at the satellite and terrestrial transmitter side, and meanwhile, the UAV-RIS is introduced for the cooperating interference elimination process. Additionally, the degrees of freedom (DoF) obtained by the proposed IM scheme are discussed in depth when the number of antennas on the satellite side is insufficient. Simulation results show that the proposed IM scheme improves the system capacity in different CSI scenarios, and the performance is better than the existing IM benchmarks without UAV-RIS, but the performance improvement is at the cost of the requirement on the elements of UAV-RIS.

In this paper, we investigate a cognitive multicast communication system aided by active reconfigurable intelligent surface (active RIS). Specifically, for an underlay spectrum sharing cognitive multicast network, a cognitive radio base station (CRBS) communicates with secondary users (SUs) assisted by an active RIS. Meanwhile, the interference to primary users (PUs) is suppressed within the constraints of the transmit power of both the CRBS and active RIS, together with the restriction of the active RIS amplitude gain. We aim at the fairness problem for maximizing the minimum signal-to-interference-plus-noise-ratio (SINR) via joint beamforming design at the CRBS and the active RIS. To cope with this problem, the optimization and deep reinforcement learning (DRL) based algorithms are proposed. Specifically, the decision variables are decoupled by the alternating optimization (AO) method and then, the non-convex problem is transformed into a solvable convex form by using the successive convex approximation (SCA), Schur complement, and penalty convex-concave procedure (PCCP) methods. Furthermore, we design an AO-based algorithm for the formulated problem. Due to the characteristics of both exploration and exploitation, the DRL-based algorithms outperform the AO-based algorithm with proper parameter settings. Meanwhile, the DRL algorithm inherits the advantages of low execution complexity. The original optimization problem is first converted into a Markov decision process (MDP) form in DRL. Due to the complex objective function and various restrictions of power/amplification gain budget and quality of service (QoS), the constraints are categorized as the switching constraints for action adjustment and performance constraints for reward function setting, respectively. Subsequently, a segmented incentive-based reward function is developed to attain higher performance on SINR. We also propose two effective deep deterministic policy gradient (DDPG)-based and twin delayed deep deterministic policy gradient (TD3)-based algorithms. Finally, the simulation results demonstrate a notable enhancement in system performance upon the introduction of active RIS compared to the case with a passive RIS and the case without using an RIS. Moreover, with appropriately configured parameters, DRL algorithms out-perform the AO-based algorithm, and notably, the TD3 algorithm is superior to the DDPG algorithm in optimization effectiveness.

Short-term load forecasting plays a significant role in the operation of power systems. Recently, deep learning has been generally employed in short-term load forecasting, primarily in the extraction of the characteristics of digital information in a single dimension without taking into account of the impact of external variables, particularly non-digital elements on load characteristics. In this paper, we propose a joint MFF-SAM-GCN to realize short-term load forecasting. First, we utilize a Bi-directional Long Short-Term Memory (Bi-LSTM) network and One-Dimensional Convolutional Neural Network (1D-CNN) in parallel connection to form a multi-feature fusion (MFF) framework, which can extract spatiotemporal correlation features of the load data. In addition, we introduce a Self-Attention Mechanism (SAM) to further enhance the feature extraction capability of the 1D-CNN network. Then with the deployment of a Graph Convolutional Network (GCN), the external non-digital features such as weather, strength, and direction of wind, etc., are extracted. Moreover, the generated weight matrices are incorporated into the load features to enhance feature recognition ability. Finally, we exploit Bayesian Optimization (BO) to find the optimal hyperparameters of the model to further improve the prediction accuracy. The simulation is taken from our proposed model and six benchmark schemes by using the bus load dataset of the Shandong Open Data Network, China. The results show that the RMSE of our proposed MFF-SAM-GCN model is 0.0284, while the SMAPE is 9.453%,the MBE is 0.025, and R-squared is 0.989, which is better than the selected three traditional machine learning methods and the three deep learning models.

To achieve ultra-reliable and low-latency communication (URLLC) and support high density of wireless connections simultaneously, the sixth-generation industrial Internet of things (6G-IIoT) necessitates an expansion of antenna arrays and broader bandwidths, which suffers from high energy consumption. To address this issue, this paper investigates a cell-free massive multiple-input multiple-output (CF-mMIMO) system and designs an iterative search-based two-stage energy efficiency (EE) optimization algorithm for the uplink communication of the system. The first stage prioritizes reliability to ensure that all users meet the URLLC requirements regarding latency and reliability. The second stage maximizes the EE under URLLC-satisfied conditions. Considering that iterative search algorithms incur a high computational overhead and variable number of iterations, we further propose a convolutional neural network architecture (RACNN) to approximate an optimal resource allocation strategy and to realize the real-time and stable output. This structure extracts deep correlations among users from the global channel features. It takes strategies of multi-task learning and weight loss adaptation to improve the model’s convergence speed. Furthermore, we employ deep transfer learning to adjust RACNN parameters to accommodate the potential of dynamic communication scenarios, thereby reducing the demand for training samples and training time overhead. Finally, the efficacy of the proposed algorithm, RACNN, and deep transfer learning is validated through experimental simulations.