Modern satellite networks are transitioning from monolithic designs to microservice architectures, introducing complex spatio-temporal patterns, varied on-board processing capabilities, and high mobility with dynamic topologies. These features require accurate prediction of inter service dependencies for optimal resource allocation and system management. To address these challenges, this paper introduces the Spatio-Temporal Graph Neural Network with Orbital contextual features (STGNN-O). This model incorporates orbital information as contextual features, processes spatial dependencies through multi-head graph attention networks, and captures temporal patterns at three different timescales to complete satellite service performence metrics prediction, which refers to forecasting key performance indicators of microservices running on satellite platforms. Also, due to the lack of real-world relavent dataset, a comprehensive satellite service benchmark dataset is created based on real-world parameters and service patterns across multiple orbital configurations. Experiments demonstrate that STGNN-O significantly outperforms state-of-the-art baselines, achieving substantial improvements in prediction accuracy. Ablation studies confirm that the integration of orbital information and multi-scale temporal features significantly contributes to prediction accuracy across all orbital regimes.

In MIMO systems, hybrid digital and analog precoding structures are gaining attention for reducing power consumption and costs compared to fully digital precoding. However, deployment expenses and circuit power consumption still pose challenges. Nonideal hardware in practical applications further impacts energy efficiency (EE), making research into high-energy-efficiency MIMO structures under such conditions essential. To tackle this issue, an exceedingly effective hybrid precoding strategy is developed, incorporating the implications of nonideal hardware characteristics. Initially, the research models the aforementioned hardware characteristics, formulates the signal model for the hybrid-structured MIMO communication system, and discerns the nonconvex problem relative to energy efficiency optimization. Subsequently, the optimization problem is effectively resolved by choosing analog radio frequency (RF) precoding and digital baseband precoding designs, including the precoding design for array gain. Furthermore, fractional programming is utilized to achieve digital precoding. An iterative optimization method is utilized to solve this problem. Finally, the mixed precoding scheme is formulated to showcase the system’s energy efficiency performance through simulation and explore the impact of imperfect hardware on the system. Additionally, to further diminish design complexity, the implementation of various channels is analyzed to propose a low-complexity method via the approximation of the equivalent channel matrix. Simulation results show that the performance of this method achieves performance comparable to existing hybrid precoding methods, when the number of antennas is sufficiently large.

Nondestructive underground target exploration by ground penetrating radar (GPR) is vital for civil infrastructure maintenance. Nevertheless, the analysis of field GPR data is both labor-intensive and time-consuming and also prone to misidentifications of underground targets with existing methods. To address these issues, a novel data-driven model, named MaskScoreNet-E, is first introduced to reduce processing time and ensure more precise and reliable instance segmentation from GPR B-scans, building upon the mask scoring R-convolutional neural network (MS R-CNN) framework with enhancements to its backbone and neck architectures. Second, aiming to rectify MaskScoreNet-E’s misidentified targets, we proposed a region-adaptive thresholding segmentation (RATS) algorithm that combines target depth-based greyscale selection and morphological operations. Experimental results on both synthetic and field datasets reveal that the MaskScoreNet-E surpasses the original MS R-CNN in terms of loss, segmentation, and bounding box metrics. The subsequent RATS algorithm significantly enhances target detection accuracy and reliability by RATS’s ability to balance diffraction wave removal and target shape preservation.

In this work, we investigate in depth the two timescale robust system energy efficiency (EE) precoding design for the multiuser dual-polarized multiple-input multiple-output (MIMO) system. To achieve good performance, low feedback overhead as well as low implementation complexity, the dual-structured linear precoding scheme is adopted, which is based on the two timescale channel state information (CSI) and the dual-polarized antenna structure. The subgrouping technique, which is based on polarization, is also utilized to divide spatially grouped users into co-polarized subgroups to further reduce channel feedback overhead. The proposed robust system EE precoding design can achieve the maximization of the worst-case system EE, with the norm-bounded channel errors of all users. This robust EE optimization is naturally decomposed into two parts. In the first part, based on the polarized spatial correlation information, the block diagonalization is utilized to design the preprocessing matrix. In the second part, based on the relationship between the mean square error and the signal-to-interference-plus-noise ratio, the linear precoding matrix can be optimized by utilizing the sign-definiteness lemma and the fractional programming technique. Specifically, the corresponding nonconvex EE fractional optimization problem is converted to a series of semidefinite programming problems, which are solved by the convex optimization method efficiently. Simulation results indicate that the proposed two timescale based dual-structured precoding has many advantages on the robust system EE performance in the dual-polarized multiuser MIMO system.

Real-time 3D rendering and interaction is important for virtual reality (VR) experimental education. Unfortunately, standard end-computing methods prohibitively escalate computational costs. Thus, reducing or distributing these requirements needs urgent attention, especially in light of the COVID-19 pandemic In this study, we design a cloud-to-end rendering and storage system for VR experimental education comprising two models: background and interactive. The cloud server renders items in the background and sends the results to an end terminal in a video stream. Interactive models are then lightweight-rendered and blended at the end terminal. An improved 3D warping and hole-filling algorithm is also proposed to improve image quality when the user’s viewpoint changes. We build three scenes to test image quality and network latency. The results show that our system can render 3D experimental education scenes with higher image quality and lower latency than any other cloud rendering systems. Our study is the first to use cloud and lightweight rendering for VR experimental education. The results demonstrate that our system provides good rendering experience without exceeding computation costs.