Edge artificial intelligence (AI) and space-ground integrated networks (SGINs) are two main usage scenarios of the sixth-generation (6G) mobile networks. Edge AI supports pervasive low-latency AI services to users, whereas SGINs provide digital services to spatial, aerial, maritime, and ground users. This article advocates the integration of the two technologies by extending edge AI to space, thereby delivering AI services to every corner of the planet. Beyond a simple combination, our novel framework, called space-ground fluid AI, leverages the predictive mobility of satellites to facilitate fluid horizontal and vertical task/model migration in the networks. This ensures non-disruptive AI service provisioning in spite of the high mobility of satellite servers. The aim of the article is to introduce the (space-ground) fluid AI technology. First, we outline the network architecture and unique characteristics of fluid AI. Then, we delve into three key components of fluid AI, i.e., fluid learning, fluid inference, and fluid model downloading. They share the common feature of coping with satellite mobility via inter-satellite and space-ground cooperation to support AI services. Finally, we discuss the considerations for the real-world deployment of fluid AI and identify further research opportunities.

Diverse Internet of Things (IoT) devices strategically deployed in remote geographical areas, spanning remote lands and offshore locations, play a crucial role in environmental surveillance. The inherent limitations in terrestrial network accessibility have necessitated the integration of these devices with satellites in 6G Non-Terrestrial Networks (NTNs), offering unparalleled flexibility and affordability through extensive coverage and robust connectivity. However, challenges such as constrained timeliness and data processing errors have posed significant obstacles to the efficacy and reliability of NTN-enabled IoT service coverage. In this article, we propose two innovative IoT service coverage approaches: the satellite-assisted manner (SAM) and the satellite-dedicated manner (SDM), to enhance IoT service coverage through IoT data acquisition and IoT-based remote sensing, respectively. Leveraging these approaches, we present a pioneering NTN-enabled surveillance architecture to facilitate efficient ecological data collection, processing, and transmission by coordinating networked communication, perception, and computing functions. Specifically, we introduce a novel multi-dimensional IoT service coverage pattern by designing three service modes—Mode C, Mode T, and Mode P—to flexibly switch, match, and collaborate, catering to specific ecological service requirements. Finally, simulation results validate that IoT service coverage can be significantly improved by 9.99% compared to the typical existing NTN IoT service pattern.

To deal with the challenges in mobility management of mega low-Earth-orbit (LEO) satellite constellations with long management delays and high signaling overheads, especially under the existing fixed and limited deployments of ground mobility management entities, the cooperative mobility management mode of medium-Earth-orbit (MEO) satellites and ground stations (GSs) has become an attractive tendency. In this paper, considering with the global non-uniform user distribution and constrained satellite storage resources, we propose a dynamic satellite-ground integrated mobility management strategy (DSG-MMS) to cope with the relative mobility among users, GSs, and satellites, which can dynamically decide the optimal GS/MEO management node with the minimal handover and migration delays. Specifically, the DSG-MMS optimization problem is modeled as distributed Markov decision processes, and a reinforcement learning (RL)-based management node selection method is presented to solve them, where each LEO satellite agent dynamically decides its own management node. To further implement the RL algorithm on the resource-limited LEO satellite agents, a novel tensor-based RL algorithm for DSG-MMS is proposed by means of the streamed low-rank tensor decomposition, where only the small-sized core tensor and factor matrices are kept and updated in the strategic optimization so as to realize low storage and computing overheads as well as fast convergence. We perform simulations for the proposed DSG-MMS with parameter configurations of actual Telesat, Kuiper, and Starlink satellite systems to evaluate the mobility management delay and overhead performances as well as the required satellite storage size for mobility management. Moreover, a case study and an architectural comparison are given to demonstrate the superiority of the proposed DSG-MMS than existing methods.

The network control plays a vital role in the mega satellite constellation (MSC) to coordinate massive network nodes to ensure the effectiveness and reliability of operations and services for future space wireless communications networks. One of the critical issues in satellite network control is how to design an optimal network control structure (ONCS) by configuring the least number of controllers to achieve efficient control interaction within a limited number of hops. Considering the wide coverage, rising capacity, and no geographical constraints of space platforms, this paper contributes to designing the ONCS by constructing an optimal space control network (SCN) to improve the temporal effectiveness of network control. Specifically, we formulate the optimal SCN construction problem from the perspective of satellite coverage factors, and apply geometric topology analysis to derive both the conditions for constructing the optimal SCN and the formulaic conclusions for SCN and MSC configurations (i.e., scale and structure). From numerical results, we investigate the tradeoff between network scale, the number of controllers, and control delays in several satellite network control scenarios, to provide guidelines for the MSC control. We also design the optimal SCN for an existing MSC system to demonstrate the effectiveness of the proposed ONCS.

The number of satellites in the current low-Earth-orbit (LEO) satellite networks continues to grow to form a mega LEO satellite constellation (MLSC). This large-scale net-working effectively improves network coverage and capacity. However, denser satellite deployment brings more severe chal-lenges to satellite handovers, such as frequent handovers, which exponentially reduce the system performance (e.g. probability of service success (PSS)) when considering inherent handover failures. To ensure service continuity, this paper focuses on the relationship between the constellation scale and handover times under seamless coverage. Specifically, we first conduct spatial geometric analysis and probabilistic analysis to derive three new conditions for MLSC seamless coverage. Then, we analyze the tradeoff relationships between the handover times and satellite coverage duration with the given constellation scale. Furthermore, we construct a mathematical relationship between the constellation scale, handover times, and PSS, indicating the tradeoff between constellation scale and system performance. The analysis effectively guides to design or adjust the constellation scale, satellite altitude, satellite coverage angle, and handover strategy according to the requirements of system performance, which has important theoretical value for MLSC system design and future research.

Integrated terrestrial-satellite networks (lTSNs) are envisioned to provide seamless broadband services through evolving terrestrial B5G/6G cellular systems and emerging mega satellite constellations. The inherent dual mobility (i.e., satellites and ubiquitous mobile users) and highly overlapped satellite footprints in such ITSNs may result in massive and frequent handovers. This increases handover delays, signaling overheads, and decision making loads, especially under existing fixed, even limited, deployments of ground mobility management functions (MMFs). Thus, a lightweight handover scheme and a mobility management architecture with dynamic MMF configurations are of great significance to guarantee service continuity and management timeliness. Specifically, this article proposes mobility management architectures with an augmented reconfigurable management plane on a higher orbit. The architectures can support flexible configurations of space-distributed MMFs for efficient mobility management together with the ground MMFs. Subsequently, to achieve rapid and unified decisions for massive handovers, we design a clustering and game-based handover decision framework including centralized and distributed decision making functions for different conditions. Based on handover procedures, simulation results validate the high performance of the proposed schemes regarding handover delays, signaling overheads, and convergence property. Notably, many academic issues are worthy of further studies under this architecture, such as higher-layer constellation design and inter-layer management structure optimization.