The rapid development of the low-altitude economy poses urgent demands for low-altitude intelligent networked systems that integrate communication, sensing, and computing. Existing fifth-generation mobile communication-advanced (5G-A) systems face limitations in coverage altitude, uplink data rate, and integrated communication and sensing coordination, making it difficult to independently support wide-area, high-dynamic low-altitude applications. The sixth-generation mobile communication (6G) satellite-terrestrial integrated network, with unified air interface and unified architecture through system-level deep integration, is expected to address the core challenges of low-altitude intelligent networking in coverage, sensing, and computing. The key use cases of low-altitude intelligent networking supported by 6G satellite-terrestrial integration for four typical scenarios: manned travel, low-altitude logistics, culture, sports, tourism, and training, and public services were analyzed. The requirements for key technical indicators were summarized, including coverage capability, positioning accuracy and integrity, communication performance, sensing capability, and computing and intelligent collaboration. Potential key technologies were analyzed, such as unified frame structure, waveform and multiple access, multi-mode access and synchronization, multi-satellite multi-beam transmission, collaborative computing and edge intelligence, spectrum sharing, and integrated communication, navigation, and sensing. A "scenario-demand-technology" framework was proposed to provide a reference for 6G satellite-terrestrial integration serving low-altitude intelligent connected applications.
Precise orbit determination is fundamental to ensuring the reliable operation of low Earth orbit satellite services such as communication, navigation, and remote sensing. Under resource-constrained space-ground communication conditions, traditional orbit determination methods that rely on ground-based processing of raw global navigation satellite system (GNSS) observations face significant challenges. To address this issue, a space-ground collaborative orbit determination low earth orbit satellite weight and prediction system architecture was proposed. Under this framework, satellites no longer merely forward raw GNSS observations but autonomously compute and transmit lightweight position and velocity coordinates. The ground system then performs efficient orbit determination and prediction based on these results, substantially reducing the demand for space-ground communication bandwidth. To enhance orbit determination accuracy, the Helmert variance component estimation method was further introduced to optimally weight the heterogeneous position and velocity information computed onboard, fully utilizing their statistical characteristics. Experimental results demonstrate that using 3-hour position and velocity coordinates as observation data, the maximum position errors for 1-day and 2-day orbit predictions are better than 43 meters and 349 meters, respectively, validating the effectiveness of the proposed method in resource-constrained scenarios.
The rapid growth of commercial space has prompted us to explore bridging public cloud virtual machines (VMs) with low earth orbit (LEO) satellite networks to enable data processing and inter-domain interconnection. However, extending static public cloud architectures to highly dynamic satellite networks presents challenges such as rapidly changing topology, limited onboard resources, and the complexity of tenant network isolation. To address these issues, we design a novel cloud gateway architecture and a geolocation-based VXLAN tunnel encapsulation mechanism that embeds the destination geographic coordinates at the source, enabling the ingress satellite to dynamically compute the optimal path, thereby reducing routing overhead and avoiding triangular routing. We also build a simulation platform based on Mininet and P4, and validate the mechanism’s effectiveness across multiple service scenarios. Experimental results show that, under dynamic constellation conditions, the system effectively ensures session continuity and tenant isolation, maintains an end-to-end average latency of around 105 ms with jitter below 70 ms, and exhibits simulation overhead that scales linearly with constellation size, demonstrating its scalability and robustness.
To support secure data transmission between satellites and underwater devices, five encryption schemes, namely Ascon, Speck, ChaCha20-Poly1305, ECC_AES, and RSA_AES, were compared. Considering the limited computing and storage resources of the target nodes, communication-command and navigation datasets were used to evaluate encryption and decryption time, CPU utilization, memory usage, and throughput on a general-purpose computing platform. The reported results indicate that Speck achieves shorter encryption times and higher encryption throughput under most tested conditions, while the schemes differ in decryption time and memory usage. These findings provide a preliminary reference for scheme selection in resource-constrained environments. Practical deployment requires further validation of authentication capabilities, implementation details, and performance on the target hardware.
As open complex giant systems, mega-constellations which carry intelligent computing services are difficult to be effectively simulated and controlled by existing simulation tools. To this end, a digital twin-enabled mega-constellations intelligent computing service simulation management system framework DT-megaAI, which builds digital prototypes containing multi-layers, and collects physical system data to drive the evolution of the digital prototype. DT-megaAI supports high-precision, low overhead simulation of three dimensions of multi-satellite collaboration for intelligent computing services through three key technologies. Finally, this paper introduces the design of DT-megaAI prototype system and illustrates its workflow and simulation ability by multi-satellite computing power aggregation simulation experiment.
