Interference cancellation based on spectrum sharing is a key solution to improve the network performance of multi-satellite systems. In this paper, we investigate a user-centric multi-satellite communication system where users are collaboratively served by multiple satellites. In this system, the integrated access and backhaul (IAB) networks, i.e., the satellite-to-user access network and the gateway-to-satellite backhaul network, are jointly considered. Our objective is to maximize the weighted sum rates (WSR) of all users by jointly optimizing the beamformers of multiple satellites and the gateway while considering the backhaul link constraints. To solve this non-convex optimization problem, a centralized algorithm is developed using the block coordinate update (BCU) method, assuming that global channel state information (CSI) is perfectly known. In addition, a low-complexity distributed algorithm based on multi-agent deep reinforcement learning (MA-DRL) is further proposed. This approach offers enhanced flexibility for implementation in multi-satellite systems and adaptability to dynamic environments. Simulation results demonstrate that the proposed distributed algorithm based on MA-DRL can achieve a performance almost equivalent to that of the centralized algorithm. It also reveals that the spatial diversity can be fully exploited through the joint beamforming of a multi-satellite system compared with a single-satellite system, while the growing number of satellites makes the backhaul constraint a more critical bottleneck for system capacity.

In satellite-assisted maritime communications, wave-induced rotational motions of unmanned surface vessels (USVs) cause severe beam misalignment with satellites, significantly degrading transmission performance. To overcome this challenge, we propose equipping the USV with a smart metasurface-based antenna to enable adaptive beamforming that dynamically compensates for USV rolling in harsh sea conditions. To facilitate effective beam tracking under long feedback delays, we design a transmission framework that ensures accurate channel state information (CSI) acquisition. Within this framework, a BLTNet-based model is developed to predict the instantaneous rolling angle of the USV, which is then used to infer the USV-to-satellite CSI for beamforming optimization. We further formulate a stochastic optimization problem to maximize the ergodic achievable rate of the uplink transmission and design a robust beamformer accordingly. Simulation results demonstrate the high accuracy of the proposed rolling angle prediction model under various settings and sea states, confirming that the corresponding robust beamforming design substantially enhances the USV-to-satellite transmission performance.

In satellite-assisted maritime communications, unmanned surface vessels (USVs) suffer from intensive rotational motions due to wave fluctuations, resulting in beam misalignment with the satellite and thus deteriorating the transmission performance severely. This paper initially proposes to leverage the smart metasurface at the USV to combat the rolling motion of the USV in hostile sea environments. Specifically, we design an uplink transmission framework for beam tracking and propose a rolling angle prediction model based on triple-layer long shortterm memory (TL-LSTM) to predict the instantaneous rolling angle of the USV. Then the USV-to-satellite uplink transmission rate is maximized accordingly through the joint beamforming design of the USV and the satellite based on the predicted channel state information (CSI). Simulation results demonstrate the robust accuracy of the proposed rolling angle prediction scheme under various sea conditions, while the corresponding beamforming optimization scheme is also significantly efficient in improving the USV-to-satellite transmission performance.

Wireless communications involving unmanned aerial vehicles (UAVs) are more vulnerable to the malicious jamming. As a promising solution, intelligent reflecting surface (IRS) can be equipped on the UAVs to achieve anti-jamming transmissions by leveraging the reconfigurable passive beamforming technique. In this paper, we study a wireless communication system where a UAV carries an IRS and acts as a mobile relay for a multi-antenna transmitter and receiver pair in the presence of a smart jammer that transmits jamming signals to the receiver and the IRS simultaneously. By jointly designing the transmit beamformer, IRS reflection phase, and UAV trajectory, we aim to maximize the average achievable rate over the entire flight with effective resistance to jamming attacks. To solve the non-convex problem, we adopt the alternating optimization (AO) algorithm and decompose the problem into two subproblems, i.e., joint optimization of the transmit beamformer and reflection phase for a specific time slot and optimization of the UAV trajectory. Simulation results show that the proposed joint optimization framework can well combat the jammer under various network settings such as changing the position of the jammer and the initial position of the UAV. The proposed algorithm has good convergence and achieves better performance than other benchmark schemes.

Spectrum sharing and interference cancellation are key solutions for multi-satellite systems to improve network performance. In this paper, we investigate a user-centric multi-satellite cell-free communication system where users are collabo-ratively served by multiple satellites. Our objective is to maximize the weighted sum rates (WSR) of all users by jointly optimizing the beamformers of multiple satellites. To solve this non-convex optimization problem, we first assume that the global channel state information (CSI) can be perfectly obtained and propose a centralized algorithm named per-satellite power constraints weighted minimum mean-square error (PSPC- WMMSE). To address practical implementation issues, we further propose a low-complexity distributed algorithm based on multi-agent deep reinforcement learning (MA-DRL). It is more flexible to be ap-plied to the multi-satellite system and is adaptive to the dynamic environment. Simulation results demonstrate that the proposed distributed algorithm can achieve almost similar performance to the centralized algorithm. Moreover, it is verified that the spatial diversity can be fully exploited via joint beamforming of the multi-satellite system compared to the single satellite system.

Accurate prediction of wind power generation is of great significance for the efficient operation of wind farms. However, traditional deep learning-based methods predict the wind power without simultaneously considering the temporal features of wind power and spatial features between variables, which leads to low prediction accuracy. This article proposes a novel wind power forecasting approach based on a graph convolution network (GCN) and a multiresolution convolution neural network (CNN), combining spatial features and temporal features. In this approach, GCN merged with maximum information coefficient (MIC) is proposed to extract the spatial correlation features between input variables, which considers the effects of multiple variables on wind power and provides interpretability for deep learning-based forecasting. On the other hand, multiresolution CNN combines multiscale convolution kernels with a new self-attention mechanism to understand local and long-term temporal features, which enables simultaneous prediction of wind power and other variables. Experiments on a real dataset prove that the proposed method is effective and accurate in short-term wind power forecasting. Comparisons with the other three state-of-the-art methods and ablation experiments also reveal the advantages of the proposed approach.