As global demand for efficient network services increases, the limitations of traditional terrestrial wireless networks are becoming more apparent. Space-air-ground integrated networks (SAGIN) have emerged as a promising solution to advance next-generation network infrastructure. However, SAGIN faces significant security challenges—including the lack of a robust security architecture, trust and data reliability issues in multi-hop transmissions, and the need to enhance network performance without compromising security—rendering traditional boundary-based defenses inadequate. Therefore, we propose SECURELINK, a decentralized zero-trust architecture tailored for SAGIN, which replaces traditional perimeter defenses with a ”never trust, always verify” approach. This approach strengthens network security and flexibility through continuous verification and adherence to the principle of least privilege. SECURELINK integrates blockchain technology to establish a multi-layered security verification and data processing scheme, addressing the dynamic and decentralized features of SAGIN. Additionally, we introduce DFRIO, a zero-trust traffic offloading method based on decentralized federated reinforcement learning and blockchain, designed to enhance network performance within maintaining security. Simulation results demonstrate that our solution significantly enhances SAGIN’s defense capability without compromising network stability, outperforming the two baseline schemes by 18.3% and 42.6%, respectively.

Recently, wireless local area networks (WLAN) with dense access points (APs) and multi-AP coordination (MAPC) have emerged as promising solutions for enabling coordinated transmission and reducing interference. Several coordination schemes have been proposed for MAPC. Among them, coordinated spatial reuse (C-SR) and coordinated beamforming (C-BF) enable simultaneous transmissions within overlapping basic service sets. However, these schemes involve trade-offs: C-SR offers lower overhead but provides limited signal-to-interference-plus-noise ratio (SINR) gain, whereas C-BF offers significant SINR improvement at the cost of increased overhead. Consequently, the optimal scheme depends on network conditions, such as AP density and user distribution. However, conventional MAPC employs only a single coordination scheme regardless of the situation. Therefore, we propose an adaptive coordination scheme selection method that estimates the expected performance of each scheme by evaluating trade-offs based on environmental information, without relying on channel state information. Simulation results confirm that our scheme selection method outperforms conventional approaches and improves system throughput. The adaptive coordination scheme selection overcomes the performance limitations of conventional MAPC, thereby accelerating the advancement of future WLANs.

6G Vehicle-to-Everything (V2X) networks require spectrum coordination while addressing user privacy concerns. Current multi-agent reinforcement learning approaches achieve coordination by sharing sensitive vehicle data, creating privacy risks. This paper introduces attention based multi agent deep reinforcement learning, a framework enabling spectrum coordination through privacy-aware deployment via dual critic architecture and multi-head attention mechanisms. Attention based multi agent deep reinforcement learning separates coordination learning (centralized training with controlled data sharing) from operational deployment (decentralized execution using only local environmental observations). Multi-head attention processes spatial, temporal, and frequency patterns to enable local-observation-based coordination without vehicle-to-vehicle sensitive data exchange. Simulations demonstrate 42% spectral efficiency improvement, 3.1 dB interference reduction, and 75% communication overhead savings while implementing a data-sharing-minimized deployment strategy. The framework provides network operators a practical solution for privacy-aware 6G V2X deployment.

The rapid proliferation of large-scale artificial intelligence (AI) applications has led to unprecedented demand for distributed, efficient, and highly scalable computing infrastructures. Traditional communication networks have evolved to support cloud-edge-end collaborative computing. However, they remain fundamentally communication-centric and lack native mechanisms for coordinating widespread computing resources. Computing power networks (CPNs) address these limitations by adopting a computation-first design that treats computing power as a unified, schedulable network resource. CPNs enable on-demand computing resource scheduling. This allows heterogeneous tasks to be dynamically matched with the most suitable cloud, edge, or terminal nodes based on real-time workload conditions, device capabilities, and network states. Moreover, CPNs achieve deep computing-network collaboration through joint optimization of routing, task placement, data exchange, and resource orchestration, providing performance gains beyond traditional decoupled architectures. Their unified and programmable control framework further supports seamless integration across heterogeneous devices, enabling flexible resource virtualization and efficient management of diverse network environments. Motivated by these unique characteristics, this survey provides a comprehensive overview of CPN technologies, their architectural foundations, resource allocation mechanisms, and their emerging role in supporting large-scale distributed AI training and inference. We highlight key challenges, discuss open research problems, and provide insights into future directions for advancing computation-centric networked intelligence.

Network device and system health management is the foundation of modern network operations and maintenance. Traditional health management methods, relying on expert identification or simple rule-based algorithms, struggle to cope with the heterogeneous networks (HNs) environment. Moreover, current state-of-the-art distributed fault diagnosis methods, which utilize specific machine learning techniques, lack multi-scale adaptivity for heterogeneous device information, resulting in unsatisfactory diagnostic accuracy for HNs. In this paper, we develop an LLM-assisted end-to-end intelligent network health management framework. The framework first proposes a multi-scale data scaling method based on unsupervised learning to address the multi-scale data problem in HNs. Secondly, we combine the semantic rule tree with the attention mechanism to propose a Multi-Scale Semanticized Anomaly Detection Model (MSADM) that generates network semantic information while detecting anomalies. Finally, we embed a chain-of-thought-based large-scale language model downstream to adaptively analyze the fault diagnosis results and create an analysis report containing detailed fault information and optimization strategies. We compare our scheme with other fault diagnosis models and demonstrate that it performs well on several metrics of network fault diagnosis.

Low Earth orbit (LEO) satellite constellations face two converging threats, adaptive jamming that learns defensive behavior and quantum computers capable of breaking today’s cryptography. To address this, we propose Q-FLAP, a unified framework that integrates quantum-secured federated learning (FL) with deep reinforcement learning (DRL), to address both challenges. Q-FLAP models anti-jamming channel selection as a distributed Markov decision process, where satellites learn policies via Deep Q-Networks while federated averaging enables coordination without centralizing sensitive data. Federated updates are secured using BB84 quantum key distribution (QKD) and lattice-based post-quantum cryptography (PQC), ensuring resilience against classical eavesdropping and harvest-now-decrypt-later attacks. Analytical results establish a steady-state throughput benchmark under adversarial jamming. Simulations show that under intelligent jamming, Q-FLAP retains 78% of nominal capacity, achieving 2.52 Gbps raw throughput and a 77.9% transmission success rate. Security overhead is minimal, 2.16% for QKD and 0.094% for PQC, yielding effective secure throughputs of 2.46 Gbps and 2.52 Gbps. Compared with baseline methods, Q-FLAP improves throughput by 2.5 times over random allocation and 1.67 times over greedy channel selection.

Intelligent reflecting surfaces (IRSs) have emerged as a promising technology for enhancing wireless communications by dynamically shaping the propagation environment using controllable passive elements. However, installation angle and position errors must be addressed for real-world IRS implementation, as neglecting these physical misalignments significantly degrades beamforming accuracy and communication quality. This study addresses this challenge by introducing the concept of IRS calibration. Calibration is the process of estimating the incident and reflection angle offsets and correcting them via phase control on the IRS side to recover optimal beamforming. Key design issues are identified by performing calibration under real-world constraints. It is proposed that decoupled angular offsets can be uniquely identified using power measurements at two receiver points. The effectiveness of the proposed method is verified via experiments in the 28 GHz band using actual IRS hardware. The experimental results confirm that the impact of angular errors is successfully reduced across various error scales and reflection directions. The findings of this study highlight the importance and feasibility of calibration as a basis for real-world IRS deployment.

Space Computing Power Networks (SCPNs), also termed as Satellite Comptuting Power Networks, as an integration of satellite networks, orbital computing, and terrestrial infrastructure, have been becoming an emerging architecture and attracting growing research attention during the past few years. Beyond meeting the differentiated intelligent communication, computing, and caching service requirements from users and terminals across space, air, ground, and sea, SCPNs hold significant importance for space exploration, earth observation, environment monitoring, remote user activities, and so on. There is no doubt that SCPNs will be the critical part of 6G to realize the ubiquitous and seamless intelligence. However, compared to traditional Terrestrial Computing Power Networks (TCPNs) and Satellite Computing Networks (SCNs), SCPNs holds the uniqueness, such as the cycled node movements, hierarchical network topology, extremely large network scalability, ubiquitous resource heterogeneity and constraints, and particular space computing environment. The system integration, protocol optimization, service orchestration, and sustainable operation of SCPN have inspired many meaningful research and projects. Considering existing survey papers mainly focus on scenarios of TCPNs carrying large-scale and complex computing tasks, this paper presents a comprehensive survey of state-of-the-art research on SCPN, covering various aspects ranging from system architecture, applications and challenges, and diversified Quality of Service (QoS) metric analysis and optimization. Finally, a number of potential future research directions have also been discussed to enlighten more innovative works.

Low Earth Orbit (LEO) satellite networks are increasingly operating in higher frequency bands such as the Ka-band to support greater bandwidth and improved service throughput. At these frequencies, both weather attenuation and Doppler shift significantly affect link reliability, signal quality, and overall throughput. This paper analyzes the individual and combined effects of weather attenuation and Doppler shift on channel quality using simulations based on real-world weather trace data. We focus on ground station selection by evaluating the impact on individual base stations. Our results show that jointly considering weather and Doppler effects leads to an improvement in accurately estimating channel capacities. This enhanced estimation enables more reliable identification of the best-performing base station, leading to improved network performance, particularly in terms of throughput, in LEO satellite communication networks.

Satellite communications constitute a significant part of 6G to enable ubiquitous connectivity, seamless services, and ultra-high mobility, but pose challenges by long propagation paths and open broadcast environment. Fortunately, Intelligent Reflecting Surface (IRS) can address above challenges by dynamically adjusting the phase, amplitude, and polarization of reflecting signals, thus finally change the wavefront and enable the signal coherent superposition at receivers, resulting in improved signal quality, which can be further adopted to promote the performance of the Integrated Communication, Navigation, and remote Sensing (ICNS). Inspired by these advantages, this article investigates the potential applications of IRS in future satellite constellations. We first provide a comprehensive review of satellite communication systems and IRS, along with their capabilities. By leveraging their respective advantages and specific characteristics, we further explore the applications of IRS deployment in satellites, including the Internet of Everything (IoE), deep space exploration, positioning and navigation, and secret communications. A case study of the IRS-secured satellite framework is presented to prevent air and space Eavesdroppers (Eves), further demonstrating its security advantages. Finally, we discuss the future perspectives of IRS-assisted satellite communication systems.

With the rapid advancement of 6G and intelligent networks, Integrated Sensing, Communication, and Computing (ISCC) has emerged as a key technology to support ubiquitous connectivity and intelligent services. However, many ISCC architectures remain performance-indicator-driven and lack task intent understanding and semantic awareness. As a result, they cannot re-prioritize objectives and constraints or coordinate cross-domain resources in real time, limiting adaptability in dynamic multi-scenario deployments. Recently, Large Language Models (LLMs) have enabled agentic reasoning and planning to interpret task intent and orchestrate cross-domain decisions in intelligent networks. To address this gap, we propose a multi-agent LLM-driven ISCC resource scheduling framework. Multiple LLM agents collaborate across functional layers, forming a perception- cognition-decision-action closed loop that enables cross-domain task decomposition, semantic reasoning, and real-time resource orchestration. This shifts ISCC scheduling from performance-indicator-driven optimization to taskand semantics-driven scheduling. We validate the framework through case studies in representative ISCC scenarios and discuss key challenges and future directions.

Distributed green data centers powered by renewable energy have garnered considerable attention owing to their potential to enhance both sustainability and disaster resilience. To interconnect these geographically dispersed sites, low-Earth orbit satellite optical communications offer a promising solution, providing low-latency and high-bandwidth connectivity. However, the efficient operation of such a system hinges on a comprehensive task scheduling strategy that addresses two critical factors: task assignment—determining which data center should process each task—and processor performance control—regulating the operational speed of each processor. The effectiveness of this strategy is deeply intertwined with data placement decisions, as the location of stored data directly influences the set of data centers eligible to execute specific tasks. Optimizing this system is particularly challenging owing to the inherent variability in two primary factors: the availability of renewable energy and the quality of satellite communication links, both of which fluctuate with weather. To address these challenges, this study proposes a framework for the joint optimization of data placement and task scheduling. The proposed approach simultaneously considers fluctuating communication and power resources to maximize user task satisfaction. The problem is formulated as a mixed-integer nonlinear programming problem, and efficient solutions are obtained through computationally efficient alternating optimization algorithms. The proposed method significantly enhances task satisfaction rates compared with conventional approaches that optimize data placement and task scheduling in isolation, particularly under heterogeneous task distributions and adverse weather conditions, demonstrating its robustness and adaptability.

This paper investigates secure multi-horizon forecasting for IoT edge energy management using a real residentialcommunity electricity load dataset. We study day-ahead joint prediction (96 steps at 15-minute resolution) of load power and storage-related states from 6 months of multivariate telemetry, given a 7-day historical window. We propose a lightweight pipeline that (i) removes weekly seasonality via slot-wise residual normalization, (ii) applies a depthwise 1-D denoising front-end to suppress sensor noise and tampering, and (iii) employs a Transformer encoder with horizon-query attention pooling for horizonspecific decoding. To harden the model, we integrate adversarial robust training in the residual domain and evaluate both whitebox and transfer attacks under L∞ and L2 bounds. Experiments against classical and deep baselines show competitive clean accuracy on load variables and substantially reduced degradation under attacks, achieving up to about 30% lower MAE at high perturbation budgets. Horizon-wise error heatmaps confirm consistent gains across the 24-hour horizon. A focused ablation study further isolates the effects of seasonal residualization, the depthwise denoising front-end, Horizon-Query decoding, and adversarial training, revealing target-dependent contributions to periodic load forecasting, storage-state modeling, high-frequency perturbation suppression, and adversarial robustness.

Although Wi-Fi 7 is able to achieve a maximum throughput of up to 30 Gbps, reliability shortfalls, high latency, and low signal transmission efficiency still exist in frontier domains such as the Industrial Internet, the Metaverse, and autonomous driving. To address these issues, the IEEE 802.11 working group initiated a new standard study, IEEE 802.11bn, also known as Ultra-High Reliability (UHR). This paper systematically surveys the UHR Task Group's research progress on the Physical (PHY) and Medium Access Control (MAC) layers, introducing specific technical enhancements including Distributed Resource Unit (DRU), Modulation and Coding Scheme (MCS), 2x Low-Density Parity-Check (2x LDPC), Cooperative Spatial Reuse (Co-SR), Cooperative Beamforming (Co-BF), Seamless Mobility Domain (SMD), High Priority Enhanced Distributed Channel Access (P-EDCA), Non-Primary Channel Access (NPCA), and Dynamic Subband Operation (DSO). Furthermore, through a survey of the IEEE 802.11 TIG Task Group and relevant literature, this paper also explores the potential use cases of AI/ML technology in the Wi-Fi standardization process. Finally, this paper proposes the challenges, open issues, and future directions for UHR standardization, contributing to the advancement of Wi-Fi.

With the rapid deployment of 5G and the advancement of 6G research, traditional network architectures face challenges in meeting the demands of massive data transmission and low-latency computing. Computing Power Networks (CPN) integrate communication and computation resources to support emerging applications efficiently. Meanwhile, the Space-Air Ground Integrated Networks (SAGIN) provides global coverage and multi-layer coordination as acore 6G architecture. This paper proposes SAGIN-CPN, a heterogeneous network architecture that combines SAGIN and CPN, and introduces TOLLM, an Adaptive Large Language Model (LLM)-Based Task Orchestration (TOLLM) scheme. TOLLM exploits the advantages of LLMs in dynamic environment perception, reasoning, and decision making. By incorporating a multi-objective optimization strategy, it enables intelligent scheduling of heterogeneous nodes in SAGIN-CPN and achieves efficient joint optimization of task latency and energy consumption. Simulation results validate the effectiveness of the proposed method in enhancing Quality of Experience (QoE). This work presents a generalizable and intelligent solution for large-scale task management in future 6G networks.

With the increasing deployment of satellites and their utilization of spectrum resources, there will be heightened competition among satellite and communication systems for these resources. To improve spectrum utilization and communications coverage, rate-splitting multiple access (RSMA) and intelligent reflecting surface (IRS) are adopted to aid the Integrated satellite-terrestrial networks (ISTN) communication. However, the transmission bottleneck of RSMA is limited to the users with the worst channel quality, and fixed relay RSMA is difficult to balance the joint optimization of public and private transmission rates for all RSMA users, which results in sub-optimal or poor performance of the ISTN communication system. To address this issue, we propose a flexible IRS-assisted grouped RSMA communication in ISTN to optimize the sum-rate, where the user can choose the satellite-IRS-user link or the satellite-user link based on channel conditions and potential obstacles. The considered optimization problem is defined as the joint problem of precoding beamforming design in satellite, passive beamforming design in IRS, rate splitting optimization, and IRS user selection. To solve this joint problem, we firstly decompose the problem into fours subproblems, which can be solved through the semidefinite relaxation and polyblock outer approximation methods, Riemannian mainfold conjugate gradient, alternating optimization and simulated annealing algorithm, respectively. Then we find the joint optimized solution via an iterative manner and analyze the computational complexity of proposed algorithms. Furthermore, numerical results show that the achievable transmission sum-rate of the proposed scheme is greater than those of three benchmark schemes, which are the IRS-assisted RSMA scheme, IRS-assisted non-orthogonal multiple access (NOMA) scheme, and IRS-assisted space-division multiple access (SDMA) scheme.

Intelligent reflecting surfaces (IRSs) have emerged as a promising solution to enhance wireless communication, particularly in high-frequency bands where signal attenuation and blockage are commonly encountered. However, owing to the passive and analog nature of IRS, generating beams that support multiple users (UEs) simultaneously is difficult, and conventional methods face limitations in scenarios requiring ultra-massive connectivity, such as 6G networks. To address this, we propose a multiaccess scheme based on IRS partitioning. The IRS is divided into multiple regions, each of which directs a beam to a specific UE. Our method jointly optimizes communication scheduling, partition positions, and phase offsets to improve performance. Numerical evaluations demonstrate that our approach significantly outperforms traditional time-division multiple access (TDMA)-based IRS in terms of transmission. Furthermore, we examined the impact of each control parameter on performance, confirming that our method mitigates destructive signal superposition at partition boundaries and enhances communication efficiency. These findings suggest that the proposed IRS partitioning scheme is a viable approach to achieving ultra-high-speed, high-capacity, and ultra-massive connectivity in next-generation wireless networks.

Many real-world problems involve hierarchical multi-objective optimization over coupled sequential decisions. A common strategy is to combine heuristic planning with deep reinforcement learning (DRL). While existing hybrid methods often show strong empirical performance, the coupled closed-loop learning dynamics between upper-tier planning and lower-tier control are typically analyzed only empirically, and explicit convergence guarantees for the integrated scheme remain limited. To address this gap, we propose a heuristic-supervised-DRL (HSD) framework that tightly couples (i) a heuristic planner for upper-tier decision-making, (ii) a DRL agent for lower-tier execution, and (iii) an online supervised predictor that serves as an adaptive bridge between planning and execution. The key novelty of HSD lies in this closed-loop architecture and its accompanying theoretical treatment. By formulating the coupled updates as a two-timescale stochastic approximation process, we show that, under standard conditions, the supervised predictor tracks a quasi-stationary regression target and the overall joint process converges almost surely to an asymptotically stable equilibrium. We further analyze robustness under approximate planning errors. As a case study, we instantiate HSD in a multi-UAV-assisted mobile edge computing system. Experimental results show that the proposed framework consistently outperforms representative baselines in key performance metrics, demonstrating both its practical effectiveness and its value as a principled framework for hierarchical optimization in dynamic environments.

Non-Terrestrial Networks (NTN) and Space-Air-Ground Integrated Networks (SAGIN) represent critical architectural foundations for 6G, enabling ubiquitous global connectivity through satellite constellations, high-altitude platforms, and terrestrial infrastructure. However, delivering computation-intensive and latency-sensitive services across these heterogeneous networks demands intelligent orchestration of distributed Mobile Edge Computing (MEC) resources across highly dynamic space, aerial, and ground layers operating over high-frequency communication links (such as Terahertz (THz), millimeter-wave (mmWave), Ka-band, and Free Space Optical (FSO) technologies). This survey systematically examines how Artificial Intelligence (AI) methods, particularly deep reinforcement learning, federated learning, multi-agent systems, and graph neural networks, address fundamental edge computing challenges unique to high-frequency SAGIN/NTN environments. We present a comprehensive taxonomy categorizing AI-driven solutions for computation offloading, resource allocation, task scheduling, service placement, load balancing, and energy optimization, while accounting for time-varying high-frequency links and satellite mobility. Through rigorous analysis of state-of-the-art literature, we demonstrate how AI techniques enable link-adaptive offloading decisions, mobility-aware service migration, and hierarchical resource management across non-terrestrial nodes and terrestrial edge servers. Finally, we outline seven critical research directions encompassing AI-native cross-layer orchestration, hierarchical offloading strategies, lightweight trustworthy AI, security and privacy preservation, energy sustainability, and standardization pathways, providing a comprehensive roadmap for next-generation intelligent SAGIN/NTN-MEC systems.

Accurate performance prediction is essential for wireless network optimization and management. Graph neural networks (GNNs) as the state-of-the-art method for network modeling, leverage the pairwise relationships in graph structure of wireless networks to propagate information. However, network performance variations are influenced by diverse environmental factors, such as the operational states of nodes within the same access point (AP) network and the conditions for nodes and links traversed by flows. The lack of many-to-many high-order information will lead to the decline of prediction accuracy. Therefore, we propose HGNet, a wireless network performance prediction model based on hypergraph learning, which constructs multi-type hyperedge groups with hyperedge characteristics and typed-weighted hypergraph message passing mechanism with temporal dynamics. Furthermore, to support privacy-preserving multi-domain deployment, we propose FedHGT, a personalized federated hypergraph structure-aware optimization strategy that addresses higher-order heterogeneity in federated hypergraph training, including the multi-category of nodes and multi-connectivity of hyperedges. The server performs adaptive weighted aggregation based on the contribution and connectivity of hyperedges. We evaluate the performance of HGNet and FedHGT in ns-3 and the results demonstrate that they outperform state-of-the-art baselines by average margins of 15.9% and 5.6%, respectively.