Low earth orbit satellite networks face severe reliability challenges in the complex space environment, where conventional routing protocols, with their reactive mechanisms, struggle to address the link instability risks induced by space environment effects. To this end, this paper proposes a resilient and intelligent routing method for low earth orbit satellite networks, driven by environmental reliability prediction. The method begins by designing a space environment reliability index that integrates predictable radiation risks, space weather disturbances, and real-time link communication quality, and subsequently formulates the end-to-end routing problem. Next, a message-passing-based model is constructed to aggregate neighborhood state information among satellites. This enables a collaborative multi-agent distributed routing algorithm, guided by environmental state predictions, to facilitate a paradigm shift in routing decisions from reactive response to proactive avoidance. Simulation results demonstrate that, compared to baseline algorithms, the proposed method can proactively avoid high-risk regions, significantly enhancing the end-to-end routing reliability of low earth orbit satellite networks in dynamic space environments.

One of the key missions of sixth-generation (6G) networks is to deploy artificial intelligence (AI) at the network edge. The widely recognized paradigm, edge inference, is envisioned to enable diverse IoT applications such as autonomous driving, industrial automation, and human activity recognition. However, its efficient deployment is hindered by two major challenges: 1) the communication bottleneck from uploading high-dimensional features under limited resources, and 2) performance degradation due to outdated channel estimation, known as channel aging. This work addresses these challenges to enable robust edge inference via efficient resource allocation. In particular, we theoretically characterize the effects of channel aging on end-to-end (E2E) performance by linking inference accuracy to a surrogate metric, termed effective discriminant gain. An efficient resource allocation strategy adaptive to channel aging is realized by leveraging the permutation invariance of a graph neural network (GNN) to maximize the proposed surrogate metric. Simulation results demonstrate the superiority of the proposed framework over benchmarks that either assume perfect CSI or employ uniform resource allocation.

Leveraging continuous solar energy harvesting at high efficiency, space data centers are envisioned as a promising platform for executing energy-intensive large language models (LLMs). Recognizing this advantage, space and AI conglomerates (e.g., SpaceX, Google) are actively investing in this vision. One key challenge, however, is the efficient distributed deployment of a large-scale LLM in a satellite network due to the limited onboard computing and communication resources. This gives rise to a placement problem that involves partitioning and mapping model components to satellites such that the fundamentally different model architecture and network topology can be reconciled to ensure low-latency token generation. To address this problem, we present the Space Network of Mixture-of-Experts (SpaceMoE) framework targeting the distributed execution of a popular mixture-of-experts (MoE) model in space. The proposed placement strategies are two-level: (1) layer placement, which assigns MoE layers to satellite subnets; and (2) intra-layer expert placement, which assigns individual experts to satellites associated with the same layer/subnet. For layer placement, we exploit the ring-like communication pattern of autoregressive inference to partition the satellite constellation along the orbiting direction into subnets arranged on a ring, each hosting one MoE layer. Based on this architecture, we formulate and solve an optimization problem for intra-layer expert placement to map experts with heterogeneous activation probabilities onto satellites. The derived strategy reveals an intuitive principle: a frequently activated expert should be mapped to a satellite on a routing path with low expected latency. Experiments over a thousand-satellite constellation show that SpaceMoE achieves at least a threefold latency reduction compared with conventional random and ablation-based placement strategies.

This paper aims to improve energy efficiency (EE) of the integrated access and backhaul (IAB) aerial-terrestrial network, facilitating rapid and adjustable network infrastructure deployment. This is challenging, as interference generated by backhaul and access links degrades network throughput, and power imbalance between these links increases overall energy consumption. To this end, we jointly optimize aerial base station (ABS) deployment, user association, and downlink power allocation for both terrestrial base station and ABSs to maximize network EE. Specifically, using fractional programming, the EE maximization problem is transformed into a subtractive-form parametric problem, and then decomposed into ABS deployment and resource allocation subproblems. A hybrid algorithm combining particle swarm optimization and simulated annealing is proposed to solve the ABS deployment subproblem, determining ABS spatial configurations and updating power allocation given fixed user association. Meanwhile, a dynamic power allocation in response to network load is designed to solve the resource allocation subproblem. Furthermore, considering the quality of service requirements of ground users and the transmit power constraints of base stations, a joint EE optimization algorithm is proposed to enhance the network EE. Simulation results validate the effectiveness of the proposed methods in improving network EE, especially in scenarios involving more deployed ABSs.

One key challenge in designing resilient large-scale wireless ad hoc networks is to understand how random node failures affect fundamental network performance. In this work, we show that both network capacity and delay scale as 0.65Θ(√(n(1-q)/log n)), where n is the total number of nodes and q is the node failure probability. The network capacity degenerates to the classical result given by P. Gupta and P. R. Kumar when q=0. Based on these results, we find that even with the same number of non-faulty nodes, a network with n nodes and node failure probability q has lower network capacity than a failure-free network with n(1-q) nodes. To compensate for the network capacity loss caused by random node failures, at least ε(n,q) nq redundant nodes are required, where ε(n,q)>1. We further prove that the optimal trade-off between network capacity and delay remains O(1) regardless of node failures, implying that high network capacity and low delay cannot be achieved simultaneously. These results demonstrate robustness against stochastic variations in wireless channels.

A distinctive function of sixth-generation (6G) networks is the integration of distributed sensing and edge artificial intelligence (AI) to enable intelligent perception of the physical world. This resultant platform, termed integrated sensing and edge AI (ISEA), is envisioned to enable a broad spectrum of Internet-of-Things (IoT) applications, including remote surgery, autonomous driving, and holographic telepresence. Recently, the communication bottleneck confronting the implementation of an ISEA system is overcome by the development of over-the-air computing (AirComp) techniques, which facilitate simultaneous access through over-the-air data feature fusion. Despite its advantages, AirComp with uncoded transmission remains vulnerable to interference. To tackle this challenge, we propose AirBreath sensing, a spectrum-efficient framework that cascades feature compression and spread spectrum to mitigate interference without bandwidth expansion. This work reveals a fundamental tradeoff between these two operations under a fixed bandwidth constraint: increasing the compression ratio may reduce sensing accuracy but allows for more aggressive interference suppression via spread spectrum, and vice versa. This tradeoff is regulated by a key variable called breathing depth, defined as the feature subspace dimension that matches the processing gain in spread spectrum. To optimally control the breathing depth, we mathematically characterize and optimize this aforementioned tradeoff by designing a tractable surrogate for sensing accuracy, measured by classification discriminant gain (DG). Experimental results on real datasets demonstrate that AirBreath sensing effectively mitigates interference in ISEA systems, and the proposed control algorithm achieves near-optimal performance as benchmarked with a brute-force search.

With the rapid development of multi-beam low earth orbit (LEO) satellite networks, multi-beam satellites have attracted extensive attention due to their flexible coverage capabilities. However, the dense deployment of satellites leads to frequent inter-satellite handovers and severe co-channel interference, particularly under non-uniform ground user terminals (GUTs) distributions, which degrade the downlink rate of GUTs. To address this issue, a coordinated beam management and resource allocation method is proposed to maximize the long-term average rate of GUTs under the inter-satellite handover frequency constraint. To address the time-varying nature of the network topology, we first employ Lyapunov optimization to transform the long-term stochastic optimization problem into three tractable short-term deterministic subproblems, i.e., beam association and inter-satellite handover, GUT access and subchannel (SC) allocation, and power allocation. Recognizing that inter-satellite handover affects the interference distribution and further affects the resource allocation strategy, a matching-based beam association and inter-satellite handover method is proposed and a virtual queue is introduced to capture and control the handover frequency. Then, a graph-based interference-aware greedy scheduling method is proposed, where co-channel interference between GUTs is mitigated through spatial isolation. Simulation results demonstrate that the proposed method significantly improves the downlink sum rate of GUTs while satisfying the inter-satellite handover frequency constraint.

Utilizing remote sensing satellites for ecological area observation to provide timely high-resolution imagery is crucial for detailed ecological analysis. However, the extensive coverage of ecological areas and the limitation that a single satellite can only cover a limited target area at any given time lead to complex scheduling conflicts of imaging resources, thereby reducing observation coverage and mission completion rates. To address these challenges, this article first proposes a multi-satellite cooperative regional target mission scheduling framework that integrates satellites, target areas and ground stations. Specifically, the target areas are segmented into grid spaces, upon which an adaptive scanning area calculation model driven by satellite attitude is developed. In addition, a strip segmentation method tailored for regional targets is designed to maximize observation efficiency while minimizing resource wastage. The communication constraints between the satellite and the ground station are also fully considered in the framework to ensure that the acquired observation data can be efficiently and reliably transmitted to the ground station. Furthermore, this article introduces a multiagent reinforcement learning-based regional mission scheduling (MSRTS) algorithm that promotes collaboration through a shared resource pool among all agents. Experimental results demonstrate that the proposed MSRTS algorithm significantly enhances the mission completion rate, with an average increase in benefits of approximately 9.72%. These findings provide robust technical support for future large-scale real-time ecological remote sensing monitoring across multiple regions.

Driven by diverse applications in the emerging low-altitude economy, modern aerial networks must inherently cater for highly heterogeneous environments, characterized by communication services under mixed service delay constraints and diverse user equipment (UE) mobility. However, such heterogeneity leads to resource allocation conflicts and imbalances, which undermine communication reliability and may result in network unavailability. To address this, we investigate resource management in uplink low-altitude heterogeneous networks. Specifically, we propose a flying access point (FAP)-coordinated multi-point packet delivery mechanism with a unified resource allocation (URA) scheme to efficiently manage spatial, frequency, and temporal resources. This includes subchannel allocation, time slot partitioning, and pilot length design. Then, we derive a lower bound (LB) on network availability (NA) and reveal that extended heterogeneity significantly degrades the LB due to: 1) resource reduction under URA; and 2) the independence in ensuring services under heterogeneity. To mitigate this degradation, we derive a closed-form condition on the required number of FAPs by relaxing the LB, thereby ensuring sufficient spatial resources to achieve the target NA. Meanwhile, we derive closed-form expressions for jointly approximating the optimal number of UEs sharing time-frequency resources and the pilot length. This optimization improves resource efficiency for NA by balancing the post-processing signal-to-noise ratio and its associated thresholds to satisfy reliability requirements under heterogeneous conditions. Numerical results validate the analysis and demonstrate that the proposed resource management strategy achieves the target NA under increased heterogeneity, thereby outperforming existing approaches.

Satellite mega-constellations (SMCs), comprising thousands of interconnected satellites, have emerged as critical infrastructure for 6G networks to achieve seamless global coverage. This article addresses two fundamental challenges in SMC operation: 1) the inherent spatial-temporal traffic heterogeneity with continuously escalating demand, and 2) the diverging Quality-of-Service (QoS) requirements for diverse traffic types requiring robust end-to-end performance guarantees. To enhance resource utilization while ensuring service differentiation, we propose a novel dual-scale traffic management framework encompassing macroscopic network-level coordination and microscopic node-level adaptation. The macroscopic component formulates a multiobjective optimization framework that strategically allocates transmission paths by simultaneously minimizing intersatellite link load disparities and end-to-end queuing delays. The microscopic component introduces an adaptive resource allocation mechanism that decomposes end-to-end QoS requirements into per-node service level agreements, employing federated learning-based traffic prediction to enable dynamic resource preallocation-based on real-time load conditions. This hybrid approach achieves load-aware resource provisioning that maximizes traffic completion rates (TCRs) while minimizing inefficient transmissions. Simulation results show our scheme outperforms the on-demand multiobjective optimization approach, improving TCRs by 17.0%-27.5% and resource utilization by 23.29-62.34% across varying loads, while reducing latency and enhancing fairness.

In data relay satellite networks (DRSNs), user satellite (US) data is transmitted to ground stations via geostationary (GEO) relay satellites (RSs). As the number of USs increases to enable real-time observation, the limited relay capacity becomes a critical bottleneck. On-board caching and processing at USs before transmission are promising approaches to alleviate relay pressure. However, constrained storage, computation and transmission (SCT) capacities pose significant challenges in meeting stringent delay and reliability requirements. This paper investigates the user capacity of a typical DRSN with integrated SCT processes, which is defined as the maximum number of USs that can be supported under both delay and reliability constraints. These constraints are quantified by delay violation probability (DVP) and data loss probability (DLP), whose expressions are difficult to derive directly due to the inherent coupling of SCT processes. To this end, tight upper bounds of DVP and DLP are derived based on a tandem queuing model with martingale-based analysis, and these bounds demonstrate exponential decay with increasing delay threshold and storage capacity. Based on these insights, a bi-level optimization problem is formulated and a two-step user capacity algorithm is proposed to efficiently obtain the user capacity under joint DVP and DLP constraints. The proposed analysis is conducted using representative DRSN parameters, and the numerical results show that the proposed methods can enhance user capacity by up to 38.2% and reduce computational complexity by around 90%. The results can provide guidance for future DRSN configuration, including satellites deployment and resources allocation.

In this paper, an air-to-ground (A2G) channel model, based on the geometry-based stochastic model (GBSM), is proposed, jointly considering the effects of high-speed motion and three-dimensional (3D) wobbles of fixed-wing unmanned aerial vehicles (UAVs). It is found that the UAV's internal vibration is periodic and narrowband, while wobbles caused by atmospheric flow are bounded and uniformly random, as verified by measurements. Accordingly, the wobbles are modeled by sinusoidal and uniform random processes, respectively, and the effect of high-speed motion is modeled by a moving factor, determined by flight altitude and radius. The channel temporal correlation function (CF) is derived based on the proposed model. Numerical results show that both the UAV's high-speed motion and 3D wobbles accelerate the decay of the CF, with the high-speed motion contributing over 80–90% of the reduction in coherence time, making it the dominant impairment in high dynamic A2G scenarios. In addition, the coherence time at lower carrier frequencies, such as L-band and S-band, is significantly longer. In addition, by comparing with existing low-altitude rotor UAV models, the effectiveness of the proposed model in such scenarios are verified.

Multi-satellite collaborative sensing and computation serves as a key technology for ensuring continuous tracking of highly maneuverable space targets. However, how to utilize limited observation information to achieve precise prediction of maneuver trajectories and thereby guide scheduling decisions in real-time to ensure tracking continuity under resource constraints remains a core challenge faced. To tackle this, we propose a Prediction-Driven Receding Horizon Scheduling framework. This framework leverages real-time trajectory prediction to drive the dynamic evolution of the task Directed Acyclic Graph structure. Building on this, a Task Collaboration and Handover-Aware Genetic Algorithm is designed, achieving joint optimization of collaborative computing decisions and satellite handover selections. The proposed scheme ensures superior tracking stability for highly maneuverable targets while reducing the perception-to-decision latency, thereby enabling future large-scale Low Earth Orbit satellites constellations to achieve more effective continuous monitoring and rapid response to sudden space threats.

In this study, we model the LEO satellite constellation with link failures as the Delta-random regular graph, where Delta (>= 4) is the number of antennas per satellite, i.e, each satellite can establish at most 0 inter-satellite links (ISLs). It is found that the critical temporary failure probability at which the constellation structure collapses is p(c)(0) = Delta-2/Delta-1 . We show that the capacity upper bound is proportional to Delta/gamma(p)L(alpha,D) when p < p(c)(0). L(alpha, D) is the average hop distance without link failures under traffic decay factor alpha is an element of (0, 1). gamma(p) < infinity is the hop stretch factor measuring the extra hops to overcome link failures. D is the diameter of constellations. When p -> 0 and alpha -> 1, our results degenerate to the capacity upper bound under all-to-all traffic model. We also show that the capacity upper bound of one-to-many traffic is root m times that of many-to-one, where m is the number of source or destination satellites. This is because, in one-to-many traffic model, the source satellite transmits identical data packets to the destination satellites. To achieve the capacity upper bound, we propose the distributed shortest-hop routing framework based on local link states (DSR-L), which uses prior geometric knowledge of the constellation structure. These results can provide effective guidance for resilience strategies regarding redundant satellites and ISL configurations.

This paper proposes a computing-communication resource interchange method to enhance network availability (NA) in low-altitude heterogeneous networks (LA-HetNets). In these networks, communication resource conflicts and imbalances, caused by extreme heterogeneity (diverse mobility, mixed delays, and hybrid transmission), and cross-regional traffic, reduce reliability and lead to unavailability. Restoring NA requires additional communication resources, yet dynamic cross-regional scheduling is limited, making locally redundant computing resources an alternative to reduce communication resource overhead. While computing resources address medium access control (MAC)-layer unreliability, physical (PHY)-layer functionalities still rely on communication resources. Thus, it remains unclear whether increasing computing resources alone can achieve target NA, especially under greater heterogeneity. We elaborate on the impact of heterogeneity on NA and show that expanding computing resources alone cannot meet target NA under high heterogeneity, as NA degrades sharply due to increased communication capability demands. To overcome this, we propose a cross-layer optimization method enabling computing-communication resource interchange to address both MAC- and PHY-layer unreliability. By reducing processing delays with computing resources while ensuring MAC-layer reliability, our method extends PHY-layer transmission delay and expands communication resources. Simulations demonstrate our approach's superiority in achieving target NA under greater heterogeneity, revealing that computing-communication resource interchange fulfills expanding communication capability demands more effectively than conventional resource overhead reduction.

As satellite networks evolve to support increasingly diverse services and artificial general intelligence (AGI), large language models (LLMs) are emerging as a critical foundation for future space systems. However, deploying LLMs on satellites is hindered by stringent constraints on onboard memory, computation, and energy. In this context, the mixture-of-experts (MoE) architecture emerges as a promising solution, leveraging sparse expert activation to enable scalable model inference. By harnessing the architectural advantages of MoE, this article provides a comprehensive overview of SpaceMoE, a new paradigm for distributed MoE inference in satellite networks. We first review recent industrial progress and emerging standardization trends that motivate the evolution toward space AGI systems. Then, we introduce the fundamentals and architectural evolution of SpaceMoE. Subsequently, we discuss three fundamental design problems in SpaceMoE, namely expert placement, expert selection, and hidden-state transmission and routing, highlighting how satellite-specific factors such as dynamic topology, battery degradation, and thermal limits fundamentally reshape their solutions. Finally, we outline promising research directions for realizing scalable, efficient, and sustainable on-orbit MoE inference in future satellite networks.

Mega satellite constellations (MSCs) based on low Earth orbit (LEO) satellites and inter-satellite links (ISLs) have become increasingly important due to the seamless coverage and high throughput. Unfortunately, the communication components of satellites are susceptible to radiation-induced single event upsets (SEUs), which lead to the failure of ISLs and the decline in network throughput. In this paper, we study the impact of SEUs on network throughput and propose MSC design algorithms to enhance the throughput. To mitigate the impact of SEUs, each satellite is equipped with low-cost mitigation techniques, under which ISLs experience different levels of impairment. Furthermore, we derive the expressions of network throughput and observe the mismatch between the traffic pattern and the network topology. Based on the expressions, we develop the MSC design algorithm to address the gap for throughput enhancement. Simulation results validate the accuracy of the theoretical results, and demonstrate that the proposed algorithm can effectively enhance the network throughput by 8.42% compared to the classical topology under the impact of SEUs.

Real-time telemetry, tracking, and command (TT&C) access scheduling is essential for ensuring stable operation and mission execution in satellite mega-constellations. However, due to visibility-constrained satellite–ground access time windows and the limited spatial availability of access antennas and multi-band frequencies, TT&C resources—access entities such as ground stations and relay satellites—exhibit strong spatio-temporal resource utilization interdependencies (e.g., antenna access contention across frequency bands and visibility conflicts) among mission access requests, thereby exacerbating the challenges of real-time TT&C mission responsiveness in satellite mega-constellations. To address the challenges, we propose a Networked-Enhanced TT&C (NETC) scheduling framework that leverages inter-satellite links (ISLs) to provide auxiliary access bands during non-visibility periods, thereby enhancing real-time responsiveness of TT&C Mission Access (TCMA) without additional infrastructure cost. Based on the NETC, we model the TCMA problem and then construct a Spatio-temporal Resource Availability Graph (SRAG) to efficiently decouple its high-dimensional solution space. Based on this, we further propose an Attention-Mechanism-Driven Deep Reinforcement Learning (AMDRL) algorithm that focuses on real-time, high-priority utilization of TT&C spatio-temporal access resources and enables dynamic awareness of evolving spatio-temporal resource utilization states, thereby supports real-time mission scheduling decisions even in highly dynamic mega-constellation environments. Extensive simulations show that our method significantly realizes faster convergence, better spatio-temporal resource utilization, and stronger real-time mission responsiveness performance—achieving 1.8 times higher mission completion efficiency and maintaining second-level mission responsiveness even under concurrent TT&C for mega-constellations.

Dynamic link failures disrupt the connectivity and geometric symmetry of the constellation structure, thereby increasing protocol overhead and degrading the effective capacity for traffic transport. The fundamental relationship between constellation size and effective capacity under protocol overhead constraints remains unclear. To this end, we define capacity scalability as the ratio of constellation capacity under non-failure conditions to protocol overhead. Specifically, if ISL states follow a two-state discrete Markov chain and the maintenance period is k ≥ 1, the upper bound of capacity scalability under the uniform traffic pattern is O(1/n), where n is the number of satellites. With perfect information about the constellation topology, the upper bound can be achieved via shortest-path routing. For any given protocol, there exists an optimal constellation deployment scale in terms of capacity scalability. When the constellation size is below this optimum scale, capacity scalability increases with constellation size, thereby improving effective capacity. Increasing the maintenance period k can improve capacity scalability, but it does not change the fact that the capacity scalability converges to zero when the constellation size exceeds the optimal scale.