In response to the increasing demand for fast response and large-scale coverage in application scenarios such as satellite data processing and vehicle remote control, this study focuses on utilizing hierarchical control and artificial intelligence technology to design a resource scheduling mechanism for space-air-ground integrated computing power networks. Air, space, and ground networks are divided into three domains, and domain controllers are deployed for resource management in the corresponding local domain. The areas are divided based on the coverage of satellites and drones to ensure that they can achieve effective service guarantees, efficient data transmission, and task processing. A multi-agent reinforcement learning-based scheduling algorithm is proposed to optimize resource utilization in space-air-ground integrated computing power networks, considering each domain controller is treated as an agent with task scheduling and resource allocation capabilities. Intelligent resource scheduling and efficient resource allocation for computing tasks are realized through collaborative learning and distributed decision-making with satisfactory delay and energy consumption constraints. Computing tasks are generated in different scenarios and processed in real time. Simulation results show that the proposed mechanism can effectively improve resource utilization and shorten task response time.

To address the challenges of insufficient computing capacity of end users and the unbalanced distribution of computing power among edge nodes in computing power networks, this study proposes an Unmanned Aerial Vehicle (UAV)-assisted Device-to-Device (D2D) edge computing solution based on incentive mechanisms. First, under constraints involving computing resources, transmission power, and the unit pricing of computing resources, a unified optimization problem is formulated to maximize system revenue. This problem aims to optimize the task offloading ratio, computing resource constraints, UAV trajectory, as well as the transmission power and unit pricing of computing resources for both UAVs and users. The Proximal Policy Optimization (PPO) algorithm is employed to establish user offloading and purchasing strategies. In addition, an iterative strategy is implemented at each time step to solve the optimization problem and obtain the optimal solution. The simulation results demonstrate that the PPO-based system revenue maximization algorithm exhibits superior convergence and improves overall system revenue compared to the baseline algorithm.

In the 6G era, a Space-Air-Ground Integrated Network (SAGIN) can provide ubiquitous coverage for Internet of Things (IoT) devices and can therefore effectively address the current inadequacies in network architecture coverage capabilities. Multi-access Edge Computing (MEC) is a crucial technology that further enhances the service capabilities of SAGIN, demonstrating significant abilities in reducing task execution latency and system energy consumption. This paper proposes an MEC-based SAGIN architecture in which satellites and multiple Unmanned Aerial Vehicles (UAVs) act as edge nodes that offers computational power in close proximity to IoT devices. Through the task segmentation of IoT devices and bandwidth allocation for UAVs and satellites, the proposed architecture intends to minimize the average network energy consumption. The problem of high network dynamics is reformulated as a Markov Decision Process (MDP), and a low-complexity adaptive decision algorithm based on Deep Deterministic Policy Gradient (DDPG) is introduced as its solution. Simulation results demonstrate that the algorithm performs well in minimizing network energy consumption and maximizing the cumulative rewards for the DDPG Agent.

Existing reinforcement learning-based traffic signal control methods primarily utilize historical and real-time traffic states at the current time step to determine the control strategy for the next time step. However, this approach results in the control strategy to lag behind the actual traffic state by one time step. To address this issue, this study proposes a traffic signal control method based on Advantage Actor Critic (A2C) using deep reinforcement learning. First, a Long Short-Term Memory (LSTM) network is designed to predict the future traffic states of a road network, to obtain the traffic state of future time steps and ensure that the formulated control strategy can respond more accurately to decision-making requirements under real-time traffic conditions. Second, a Kalman filter is designed to fuse collected historical traffic state data with the future traffic state data predicted by the LSTM, to improve the accuracy and robustness of the data being input into the deep reinforcement learning model. Additionally, a bidirectional LSTM-integrated A2C algorithm is proposed that allows the deep reinforcement learning model to fully capture the time-dependent relationships within traffic flow and achieve more efficient and stable traffic signal control decisions. Finally, simulations conducted on the Simulation of Urban Mobility (SUMO) platform demonstrate that the proposed method achieves superior traffic signal control efficiency under both low-peak, off-peak and peak traffic conditions compared to traditional traffic signal control methods and deep reinforcement learning A2C-based traffic signal control method.

In multitier space-air-ground computing power networks, the complexity and non-stationary characteristics of load data in a digital twin logical testbed hinder load data prediction accuracy. This study proposes the GCSG model, which integrates the Gramian Angular Field (GAF) transformation, a Convolutional Neural Network (CNN), a Squeeze-and-Excitation Network (SENet) with a channel attention mechanism, and a Gated Recurrent Unit (GRU) to achieve efficient feature fusion and precise prediction, to address this issue. The GAF transforms one-dimensional load data into two-dimensional images, enabling the CNN to extract local features. SENet optimizes feature importance through attention mechanisms, while GRU captures temporal dependencies, ensuring robust feature integration. In addition, the model employs the Variational Mode Decomposition (VMD) enhanced by the Sparrow Search Algorithm (SSA) to stabilize the load data and further improve prediction performance. Experimental results demonstrate that the GCSG model achieves superior prediction accuracy and stability across varying data lengths and excels in multistep prediction tasks. Thus, the GCSG model significantly enhances load data prediction accuracy, offering a powerful solution for load forecasting in digital twin systems within space-air-ground computing power networks.

Satellite networks have wide coverage, strong mobility, and ultralow power consumption, which allow them to act as an extension to ground communication networks, thereby promoting the construction of integrated space-ground networks. However, the opening and popularization of satellite services have increased network traffic and made it more complex, making their management and service scheduling challenging. Thus, designing an efficient network traffic classification method and allocating reasonable computing resources to different types of satellite network traffic have become critical to alleviating the pressure on satellite networks. Traditional network traffic classification methods based on ports, payloads, statistics, and behavior have issues concerning effectiveness and privacy, making them inadequate for complex network services. Various technologies are widely applied in the development of large models. Therefore, to enhance the operational efficiency of satellite networks and optimize their computing power, this study proposes a network traffic classification method based on the Global Perception Module (GPM) and ViT (Vision Transformer) model. This method transforms network traffic data into grayscale images and extracts features to fully capture global and local information. The processed data are then input into the ViT model, which leverages its multihead attention mechanism to extract data correlation information and enhance classification capability. Experimental results indicate that the accuracy of the GPM-ViT model reaches 97.86%, which is a significant improvement over that of baseline models.