Issue Ebookcover storyIssue CoverIssue Contents
Online FirstList of IssuesDocument Types
- Abstract:As the core resource for wireless communications and sensing, the electromagnetic spectrum is undergoing a paradigm shift from passive management to active cognition under the impetus of intelligent technologies. Electromagnetic-spectrum embodied intelligence deeply couples the agent with the electromagnetic environment to achieve autonomous sensing, cognition, and control of spectrum resources. This paper provides a systematic review of the theoretical framework, enabling technologies, and application prospects of electromagnetic-spectrum embodied intelligence. Starting from embodied-cognition theory, it argues for the legitimacy of “electromagnetic organs” as components of the body of an agent, reconceptualizes the electromagnetic spectrum as a key environmental dimension on par with three-dimensional physical space, constructs a mathematical model of the electromagnetic embodied closed loop, and reveals the essence that intelligence emerges from the dynamic coupling between body and environment. Building on a three-layer “sensing-cognition-action” architecture, we analyze core enabling technologies: at the electromagnetic sensing layer, across the signal, spatial, and multi-domain-fusion dimensions, we explain how software-defined radio, cognitive radar, cooperative spectrum sensing, and multimodal fusion enable a leap from physical signals to semantic understanding; at the cognition and decision-making layer, for both platform-constrained and multi-platform collaborative scenarios, we examine how lightweight reinforcement learning, online meta-learning, multi-agent collaborative decision-making, and game-theoretic optimization support autonomous decision-making and collective coordination in dynamic, adversarial environments; at the action-and-feedback layer, from electromagnetic-environment interaction to embodied-platform reconfiguration, we discuss how dynamic spectrum access, jamming and counter-jamming, waveform reconfiguration, and hardware reconfigurable technology translate cognitive policies into electromagnetic behaviors and close the loop. We summarize application practices and advantages in representative scenarios such as 6G mobile communications, intelligent electromagnetic system of LEO mega-constellations spectrum governance for smart cities, and electromagnetic operations on intelligent battlefields. In view of current challenges—including real-time constraints, sample efficiency, safety and robustness, and interpretability—we outline future trends involving standardization, ecosystem development, cross-disciplinary theoretical integration, and large-scale engineering deployment. The study offers a systematic theoretical reference and a technical roadmap for intelligent spectrum management, contributing to the modernization of spectrum-governance systems and supporting the construction of national information infrastructure.Keywords:electromagnetic spectrum;embodied artificial intelligence;electromagnetic organ;reinforcement learning;multi-agent systems431|293|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:Embodied intelligent navigation enables UAV autonomy in complex low-altitude scenarios, yet existing research often overlooks dynamic electromagnetic interference. To mitigate this, it is essential to extend navigation from the physical geometric space to the electromagnetic spectrum, achieving synergistic avoidance of obstacles and interference. To address the state-space explosion caused by coupling these domains, this paper introduces the middleware paradigm into electromagnetic embodied intelligence and proposes a UAV Embodied Anti-interference Navigation Middleware. Embedded between the mission planning layer and the flight control execution layer, it functions as a pluggable service module. It utilizes a RMDP (Residual Markov Decision Process) to decouple the optimization problem into a baseline heading and an anti-interference residual heading. By limiting the policy to learning local continuous corrections, this design significantly reduces decision complexity. Consequently, it ensures global convergence while effectively suppressing local electromagnetic risks. We further develop a RRL (Residual Reinforcement Learning) algorithm incorporating a predictive state encoding mechanism to infer wide-area interference situations from sparse observations, enabling proactive interference avoidance. Additionally, a collaborative training paradigm combining offline pre-training with controlled online fine-tuning is adopted to ensure rapid adaptation and cross-scenario generalization in unknown environments. High-fidelity simulations based on real urban maps and authentic onboard equipment specifications demonstrate that the method achieves a 93.6% task success rate with a 49.7 ms decision latency. Integration tests further verify that the architecture is upwardly compatible with heterogeneous global planners and downwardly compliant with constraints, offering significant reusability advantages.Keywords:electromagnetic embodied intelligence;unmanned aerial vehicle navigation;anti-jamming navigation middleware;Residual Markov Decision Process;Residual Reinforcement Learning185|150|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:Passive tracking of uncooperative maneuvering radiation sources by a single unmanned aerial vehicle (UAV) using only received signal strength indicator (RSSI) measurements is of considerable strategic and engineering value for missions such as electronic reconnaissance, spectrum monitoring, and emergency search and rescue. However, existing RSSI localization methods rely heavily on environmental propagation priors and treat parameter estimation, state filtering, and path planning as isolated serial processes, leaving them unable to handle dynamic scenarios in which core propagation parameters such as transmit power and path loss exponent are unknown and the target keeps maneuvering. The inherent low spatial resolution of scalar RSSI from a single channel further intensifies the structural conflict between exploration and exploitation. To address these challenges, this paper proposes an embodied intelligence-based closed-loop tracking method for a single UAV, which adopts structured uncertainty metrics as a unified carrier of cross-layer information flow and deeply couples perception, cognition, and decision-making through bidirectional propagation of uncertainty, forming a cooperative tracking architecture driven by closed-loop information flow. At the perception layer, a parameter prior inference network is built to extract the empirical distribution of propagation parameters from historical spatio-temporal observation sequences, providing data-driven initialization for the subsequent Bayesian inference. At the cognition layer, a dual-chain decoupled inference mechanism based on Rao-Blackwellized conditional decomposition is designed for the joint estimation of unknown propagation parameters and maneuvering states, in which the outer particle filter estimates large-scale fading parameters online while the inner adaptive interacting multiple model filter accurately infers maneuvering states of the target. A particle revival mechanism guided by effective sample size monitoring is further introduced together with a belief inheritance mechanism to prevent parameter resampling from disrupting the continuity of state tracking. At the decision layer, to mitigate the exploration-exploitation conflict noted above, a hierarchical planning strategy fusing information gain with model predictive control is constructed, and a learnable ranking gating network is introduced to perform real-time utility evaluation over candidate strategies including tracking, probing, and conservative options, allowing the UAV to adaptively transition from broad-area exploration to tight escort flying according to belief uncertainty. Simulation results show that the proposed method outperforms the compared baseline approaches across all core metrics under cold-start conditions with completely unknown propagation parameters, and the tracking performance under the mode without manual calibration priors approaches and in certain metrics surpasses the ideal performance of several baselines operating with known parameters, which verifies the adaptability and robustness of the proposed architecture under degraded sensing conditions.Keywords:embodied intelligence;target navigation;RSSI;active path planning;information-driven decision190|101|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
Radio Map Construction and Non-Cooperative Radiation Source Localization Based on Agents Interaction
Abstract:Radio map construction and non-cooperative emitter localization are pivotal tasks for accurate perception and dynamic control of the electromagnetic environment, characterized by high coupling and mutual constraints. Drawing on cognitive learning and embodied intelligence theories, this paper proposes a unified dual-task framework for radio map construction and emitter localization featuring multi-agent collaborative interaction and endogenous intelligence, thereby establishing a “perception-cognition-action” feedback loop. First, leveraging the concept of crowdsourced probabilistic consensus, we integrate Gaussian process regression (GPR) with 2D-positional encoding to transform discrete sparse sampling data from multiple agent into spatial attention maps. These maps encompass continuous signal field predictive distributions, geometric topologies, and probability estimates of emitter spatial presence, establishing a consensus perceptual prior regarding the target emitters. Building upon this, a consistent joint representation for emitters is constructed within a unified multi-agent semantic space. A multi-head self-attention mechanism is then utilized to design task-driven and target-driven agent interaction strategies, enabling the simultaneous execution of both tasks. Furthermore, a dual-task information interaction feedback strategy based on proximal policy optimization (PPO) is designed to achieve self-organization and self-learning, forming an embodied behavioral loop of perception, cognition, and interactive decision-making. Simulation results validate the feasibility and superiority of the proposed scheme. Under a 5% sparse sampling condition, the synergistic optimization of both tasks achieves accuracy improvements of 18.2% and 43.5% for radio map construction and emitter localization, respectively, compared to independent task execution. Specifically, the RMSE (Root Mean Square Error) of radio map construction improves by 60.95% and 32.55% over Kriging and full convolutional auto-encoder, respectively. Meanwhile, the localization error is reduced by 65.7% compared to the RSS-based baseline, maintaining a stable relative accuracy within 1.2% even when extended to a 10 km×10 km large-scale scenario.Keywords:agent interaction;embodied intelligence;consensus perception;radio map;non-cooperative radiation source localization258|99|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16- Abstract:In dense urban environments, blockage by high-rise buildings, concurrent radiation from multiple wireless systems, and the superposition of multi-source interference make the electromagnetic environment highly time-varying and spatially heterogeneous. Traditional radio maps, which are constructed offline and updated at a low frequency, cannot reflect the real channel state in time. As a result, they are unable to meet the requirements of dynamic spectrum management, wireless resource scheduling, and low-altitude communication support in terms of both accuracy and timeliness. This challenge is more pronounced in complex mission scenarios, where online sensing and rapid decision-making are especially important. To address this issue, this paper proposes an uncertainty-driven active sensing method for UAV (Unmanned Aerial Vehicles)-based spectrum mapping. First, the online radio map updating and path planning processes are jointly modeled as a Markov decision process. The state explicitly includes statistical measures of map uncertainty, UAV position, flight time, target distance, and mission progress, so as to improve the policy’s ability to represent environmental changes and mission stages. Second, a pre-trained U-Net is employed to reconstruct the SINR (Signal to Interference plus Noise Ratio) radio map online and to output pixel-level uncertainty estimates. The reduction in uncertainty is used to characterize information gain, which guides the UAV to sense high-value unknown regions with priority. Furthermore, a FiLM-D3QN decision network based on feature-wise linear modulation is designed. It uses uncertainty and mission-progress conditions to dynamically modulate intermediate features and value estimation, thereby achieving an adaptive balance between information acquisition and trajectory efficiency. Simulation results show that, in urban scenarios with dense interference, the proposed method effectively reduces map reconstruction error while maintaining an arrival rate above 90%. The RMSE (Root Mean Square Error) is reduced by about 7.1% compared with the classical IPP method and by about 9.8% compared with the original D3QN. The proposed method also achieves a shorter average episode length. These results verify its effectiveness for online radio map updating and active sensing in complex urban electromagnetic environments.Keywords:urban electromagnetic environment;unmanned aerial vehicle (UAV);radio map;active sensing;uncertainty estimation;deep reinforcement learning237|113|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:Embodied intelligent agents often fail in long-horizon tasks due to perceptual uncertainty combined with complex semantic structure in real environments. Under electromagnetic spectrum constraints, spectrum occupancy and interference can cause missing observations and link latency, further weakening the robustness of long-horizon decision-making. In this context, the agent should not only trigger help-seeking at critical moments, but also dynamically assess when to ask questions based on time-varying risk, and continuously verify residual uncertainty after receiving assistance. Most existing active querying mechanisms rely on static uncertainty criteria and assume immediate resumption of execution after a query, lacking the ability to jointly model environmental semantic risk and internal cognitive state, as well as to perform continual verification during inference. To address these limitations, we propose a metacognitive and multimodal embodied decision framework (M²ED). First, we build a task-oriented semantic environment complexity extraction module on top of a general-purpose vision-language model, which converts the current observation and dynamic instructions into structured measures of entity density, relational complexity, and descriptive entropy, forming an interpretable semantic complexity vector. Second, we design a metacognitive uncertainty-guided querying controller (MUQC), which takes the semantic complexity vector and policy perplexity as joint inputs. The controller is learned via a two-stage procedure, consisting of offline pretraining with soft-label supervision and online fine-tuning with proximal policy optimization (PPO), enabling adaptive query timing with fewer queries in simple scenarios and more queries in complex ones. Finally, we introduce an iterative metacognitive verification loop (IMVL) at inference time. Once a query is triggered, the agent pauses task execution and performs multi-round constrained perception and continuous risk calibration, resuming action execution only when both the policy perplexity and the risk score induced by the semantic complexity vector fall into a low-risk regime, thereby suppressing the accumulation and propagation of residual uncertainty over long horizons. Experiments on unseen environments in the action learning from realistic environments and directives (ALFRED) embodied benchmark show that M²ED improves task success to 19.6% while maintaining efficient interaction, limiting the average number of questions to 4.9, and demonstrating strong cross-scene generalization and long-horizon robustness.Keywords:embodied agent;vision-language model;active help-seeking;reinforcement learning161|93|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:With the rapid development of the low-altitude economy, unmanned aerial vehicles (UAVs) have been increasingly applied in scenarios such as environmental sensing, intelligent inspection, emergency rescue, and intelligent transportation. However, due to the lack of continuous environmental perception and real-time decision-making capabilities, traditional communication systems find it difficult to simultaneously achieve efficient beam tracking and low data transmission latency in highly mobile and dynamically changing environments. To address the demanding data offloading and beam tracking requirements in high-dynamic low-altitude UAV scenarios, this paper proposes an embodied-intelligence-driven electromagnetic spectrum resource allocation method. Specifically, by leveraging integrated sensing and communication (ISAC) technology, an embodied intelligent base station (BS) architecture with a closed-loop “perception-cognition-execution (PCE)” capability is constructed, where the BS is modeled as an intelligent agent capable of continuously interacting with the environment and evolving online. The proposed architecture utilizes sensing functions to obtain UAV location information in real time and assists beam prediction and resource allocation based on sensing results, thereby effectively reducing the beam alignment overhead caused by conventional pilot-based searching. Furthermore, this paper establishes a quantitative mapping relationship between sensing accuracy and sector width in the beam alignment process, enabling unified modeling of the impact of sensing performance on communication processes. The spectrum resource allocation problem in low-altitude UAV data offloading is then formulated as a multi-objective markov decision process (MOMDP) to jointly optimize sensing accuracy and data offloading latency. On this basis, the performance tradeoff caused by spectrum sharing between sensing and communication is further analyzed, enabling the resource allocation strategy to dynamically adjust the proportion of sensing and communication resources according to environmental variations, thereby improving overall system efficiency and task completion timeliness. To solve this problem, a multi-objective reinforcement-supervised hybrid online learning (MO-RSOL) algorithm is proposed. The proposed algorithm generates multi-objective experience data through environmental interaction, combines supervised learning mechanisms to achieve stable policy network updates, and continuously expands the policy performance boundary through a Pareto-coverage-oriented dataset maintenance mechanism, thereby enhancing the online adaptability and continuous evolution capability of the system in dynamic environments. Simulation results demonstrate that the proposed embodied intelligent BS architecture can reduce data offloading latency by approximately 21.7% compared with traditional non-embodied intelligent BS designs. Moreover, compared with existing baseline algorithms, the proposed MO-RSOL algorithm improves system latency and sensing performance by approximately 21% and 69%, respectively, while effectively adapting to complex scenarios involving UAV quantity variations, traffic load fluctuations, and dynamic environmental changes. These results validate the effectiveness and superiority of the proposed method in low-altitude intelligent spectrum management and high-dynamic wireless resource allocation, while also providing new research insights and theoretical references for the design of embodied intelligent communication systems in future 6G low-altitude intelligent networks.Keywords:data offloading;beam tracking;embodied intelligence;integrated sensing and communication (ISAC);resource allocation198|107|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
Electromagnetic Spectrum embodied Intelligent techniques
- Abstract:Hydrogen-terminated diamond-induced two-dimensional hole gas is an important pathway for achieving semiconductor behavior in diamond. Mechanical polishing, as a key process in the preparation of diamond substrates, inevitably introduces a subsurface damage layer, leading to surface lattice distortion and increased defect density. During selective etching in hydrogen plasma treatment, it causes deterioration of surface morphology and an increase in roughness, significantly suppressing the transport properties of the two-dimensional hole gas and limiting the electrical performance and device applications of hydrogen-terminated diamond. This study focuses on (001)-oriented chemical vapor deposition single-crystal diamond substrates. By controlling the inductively coupled plasma (ICP) etching time, the subsurface damage layer can be removed, and the effects of ICP etching time on diamond surface morphology, damage layer structure, and hydrogen-terminated electrical properties were systematically investigated. The results show that with increased ICP etching time, the etching pits and scratches on the hydrogen-treated sample surfaces gradually disappear, and surface deterioration is effectively suppressed. Transmission electron microscopy reveals that polished diamond has a subsurface damage layer of about 80 nm, which can be almost completely removed after 15 minutes of ICP etching, leaving only a surface reconstruction layer of approximately 1.5 nm. Moreover, as the damage layer is gradually removed, the carrier concentration of hydrogen-terminated diamond decreases from
to , while mobility increases from to . Further homoepitaxial growth on the damage layer-removed samples produced a low-defect, high-purity diamond epitaxial layer, with mobility further increased to . This study clarifies the regulation of hydrogen-terminated diamond electrical performance through ICP etching removal of the subsurface damage layer and provides theoretical and experimental references for optimizing the surface quality and electrical characteristics of hydrogen-terminated diamond.Keywords:single-crystal diamond;homoepitaxy;hydrogen terminal;inductively coupled plasma etching;damage layer;electrical properties203|100|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16 - Abstract:Air-sea cross-medium communication is a crucial technology for enabling high-speed air-sea cross-domain interconnection. Owing to its excellent oceanic penetration characteristics, the blue-green light band (450~550 nm) has emerged as an ideal transmission window for underwater wireless optical communication links, providing vital support for high-speed cross-medium communication. However, subjected to the nonlinear coupling effects of the atmosphere, random sea surface waves, and seawater scattering and absorption, the cross-medium channel exhibits significant non-stationarity and strong time-varying characteristics. In such rapidly time-varying channel environments, traditional fixed coding and modulation strategies, which rely on quasi-static channel assumptions, suffer from degraded system error performance due to mismatches with the actual channel conditions. To address this issue, this paper proposes a transmission scheme based on dynamic channel feature learning for air-sea cross-medium communication. By constructing a closed-loop feedback mechanism of “periodic probing—dynamic learning—reconstruction”, the coding structure can be dynamically optimized in response to varying sea states. Specifically, the transmitter periodically transmits pre-designed pilot sequences. The receiver utilizes these pilot signals for channel estimation, compiles real-time statistics on the instantaneous characteristics of the current channel, and subsequently generates a polar code construction scheme that optimally matches the prevailing sea conditions. This construction information is then fed back to the transmitter via a feedback link. Based on this information, the transmitter executes code reconstruction and applies adaptive polar coding to the information sequence to be transmitted. This closed-loop feedback process operates periodically, ensuring that the code reconstruction closely tracks the non-stationary variations of the cross-medium time-varying channel, thereby overcoming the mismatch between fixed coding strategies and the actual physical channel. To verify the performance of the proposed system, a cross-medium communication channel model encompassing the atmosphere, the air-sea interface, and the seawater medium was established. Three typical water types—Jerlov IB, Jerlov II, and Jerlov Ⅲ—were selected to simulate oceanic environments with weak to strong absorption and scattering effects, respectively. Simulation results demonstrate that the proposed dynamic channel feature learning strategy can effectively adapt to variations in the seawater medium. Compared to uncoded systems, applying polar coding under the same Jerlov water type yields a significant coding gain, with the bit error rate (BER) markedly reduced. Furthermore, compared to the fixed polar code construction strategy adopted in the 3GPP-5G standard, the proposed dynamic channel feature learning strategy achieves substantial performance improvements across all three Jerlov water conditions. Notably, the performance gain increases as water turbidity rises and channel conditions deteriorate. These results validate that adaptive coded transmission based on dynamic channel feature learning can perceive and adapt to the dynamic changes of the seawater medium, providing a solid theoretical foundation and technical support for the development of highly reliable air-sea cross-medium adaptive communication systems.Keywords:air-sea cross-medium communication;polar codes;channel characteristic dynamic learning;channel estimation;closed-loop feedback;adaptive coding80|68|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:With the rapid advancement of 5G-A, integrated sensing and communication (ISAC), and the low-altitude economy, computing resources are accelerating their extension toward metropolitan edges and space-air-ground dimensions. However, existing computing power governance systems face severe challenges, including scenario fragmentation, difficulties in measuring heterogeneous resource metrics, high dynamic uncertainty of socialized nodes, and a granularity imbalance in multi-scale scheduling, leading to a significant amount of fragmented computing power forming “computing power black holes”. To address these issues, this paper proposes a novel computing infrastructure architecture: computing star network (CSN). CSN aims to break the ownership boundaries of physical facilities and construct a space-air-ground integrated computing resource pool covering metropolitan, regional, low-altitude, and space dimensions. At the physical level, CSN adopts a dual-element architecture of “macro-core steady-state support and edge micro-core agile compensation”, which ensures a deterministic foundation for regional-scale large computing power while achieving ubiquitous access for massive metropolitan and low-altitude scattered nodes. At the logical level, a four-layer bottom-up architecture is designed: the omni-spectrum resource abstraction layer utilizes digital twins and normalized computing units (NCU) to mask hardware heterogeneity and embeds a survival analysis engine to predict node churn risks; the probabilistic semantic control layer introduces a semantic-triggered update mechanism based on age of information (AoI) to minimize control signaling overhead while ensuring state timeliness; the multi-scale intent orchestration layer translates business intents via natural language processing (NLP) and deploys a core scheduling engine; and the ubiquitous trust security layer ensures the security of socialized node access through continuous authentication and sandbox isolation. In terms of core mathematical modeling, this paper constructs a high-dimensional dual feature space. On the supply side, a node feature vector is established encompassing four dimensions: computing attributes, network perception, spatio-temporal survival, and semantic trust, specifically introducing the Weibull distribution to describe the conditional retention probability of nodes. On the demand side, computing tasks are modeled as feature vectors containing workload, QoS constraints, semantic attributes, and decomposability identifiers. Building upon this model, a scheduling algorithm utilizing a graph attention mechanism (CSN-GAM) is proposed. This algorithm performs rapid pruning through three-level hard-constraint gating (security, survival, and latency) and utilizes multi-head attention scoring to achieve precise matching between tasks and resources. Simulation experiments in a 100 km × 100 km metropolitan-scale scenario demonstrate that CSN-GAM maintains a task completion rate of over 85% under a high concurrency pressure of 500 tasks/s, significantly outperforming traditional Cloud-Only and Greedy-Edge strategies. Meanwhile, the mechanism stabilizes the average service response latency at approximately 50 ms, achieving rational offloading of heterogeneous tasks. Experimental data further confirm that the AoI-based semantic triggering mechanism saves approximately 60% of signaling overhead. Furthermore, in high-speed mobility scenarios of 30 m/s, the system maintains a task success rate of over 70% through proactive re-routing guided by the survival analysis model. In conclusion, through normalized measurement, probabilistic modeling, and multi-scale coordination, the computing star network provides a feasible technical path for the unified management and efficient scheduling of omni-spectrum heterogeneous computing power, holding profound theoretical implications and application value for the development of future space-air-ground integrated computing networks.Keywords:computing star network;omni-spectrum resources;survival analysis;graph attention mechanism;low-altitude economy137|72|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:The large-scale deployment of internet of vehicles (IoV) applications such as autonomous driving, high-definition map updates and in-vehicle entertainment has triggered an explosive growth in data volume, imposing higher requirements on cache task scheduling, content processing and recommendation, as well as network resource allocation. As a key technology to alleviate the network load of IoV and reduce the end-to-end delay of content acquisition, cooperative caching’s performance highly relies on the in-depth collaboration of three core links: vehicle matching, content recommendation and resource optimization. However, existing schemes generally suffer from poor stability of vehicle matching, insufficient recommendation adaptability and low resource optimization efficiency, making it difficult to adapt to the demands of the highly dynamic IoV environment. To address these challenges, this paper conducts research from three dimensions of communication matching, content adaptation and resource optimization, and proposes an intelligent cooperative caching strategy for IoV. Firstly, aiming at the problems of poor communication link stability and frequent handovers caused by the high mobility of vehicles, a short-term lane prediction-driven vehicle cooperative matching (SLPVCM) algorithm is designed. This algorithm comprehensively explores multi-dimensional features including spatiotemporal correlation of lane flow, vehicle communication rate and node resource occupancy rate, to realize the collaborative matching between demand vehicles, service vehicles and roadside units, thus establishing a stable communication foundation for content recommendation and resource optimization. Secondly, in view of the insufficient adaptability of traditional cache recommendation strategies to new and existing user groups and the inadequate optimization of personalized indicators, a hybrid recommendation based on collaborative filtering and content features (HRCF2) algorithm is constructed. For new users, a dual-dimensional recommendation strategy of popularity and diversity is designed to solve the cold-start problem; for existing users, the recommendation weights of collaborative filtering and content features are dynamically adjusted according to historical behaviors, which effectively improves the matching degree between cached content and user demands. Compared with the benchmark algorithms, this algorithm increases the recommendation hit rate by 21.11% and improves recommendation indicators such as F1-score by 20.94%, while reducing redundant data transmission and the dependence on remote cloud resources. Finally, facing the dynamic environmental state formed by collaborative matching relationships and content recommendation distribution, a soft Actor-Critic based vehicle caching tasks and resources optimization (SVCTRO) algorithm is proposed. By maximizing the policy entropy, this algorithm achieves adaptive and dynamic decision-making for cache task decomposition and bandwidth resource allocation, reducing the cache task costs by 5.80% and 13.27% respectively compared with the twin delayed deep deterministic policy gradient (TD3) and deep deterministic policy gradient (DDPG) algorithms. Simulation experimental results show that the proposed intelligent cooperative caching strategy outperforms the existing benchmark algorithms in terms of communication efficiency, recommendation indicators and resource scheduling economy, and can effectively adapt to the highly dynamic characteristics of IoV. It provides an effective theoretical support for the design of cooperative caching systems in the vehicle-road-cloud integration scenario.Keywords:internet of vehicles;cooperative caching;recommendation algorithm;deep reinforcement learning;resource optimization237|323|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:Reconfigurable intelligent surface (RIS) assisted wireless communication technology has been rapidly developing in recent years, which is regarded by the industry as one of the key technologies for 6G communications. RIS is a surface composed of a large number of controllable reflecting elements, which can dynamically adjust the amplitude and phase of these elements according to different radio electromagnetic environments and communication requirements, providing new dimensions and capabilities for physical layer security. Nevertheless, traditional RIS suffer from significant drawbacks such as and partial coverage and “double fading”, which hinder its flexible deployment in practical scenarios and severely limit the performance enhancement of communication systems. The active simultaneously transmitting and reflecting reconfigurable intelligent surface (STAR-RIS) can amplify the incident signal and provide full-space coverage for communication, which can overcome these shortcomings, holding greater potential for enhancing wireless secure communication systems. Moreover, cooperative jamming can interfere with eavesdroppers by transmitting jamming signals through independent collaborative nodes. Therefore, through the combination of the active STAR-RIS and cooperative jamming, we explore an active STAR-RIS-assisted downlink secure transmission system under cooperative jamming in this paper, establishing a joint optimization problem for sum secrecy rate maximization by designing beamforming vectors of the base station and the jammer, as well as the phase shifts and amplitudes of the active STAR-RIS. To solve the resulting non-convex problem, the beamforming vectors, the phase shifts and amplitudes of the active STAR-RIS in the original problem are decoupled by utilizing an alternating optimization algorithm. Firstly, with the phase shifts and amplitudes of the active STAR-RIS fixed, the transmit beamforming vectors of the base station and the jammer are optimized by employing the semi-definite relaxation algorithm. Next, with the transmit beamforming vectors of the base station and the jammer fixed, the phase shifts and amplitudes of the active STAR-RIS are optimized by employing the successive convex approximation algorithm. Finally, an iterative alternating optimization approach is employed to jointly optimize the transmit beamforming vectors of the base station and the jammer, as well as the phase shifts and amplitudes of the active STAR-RIS. Simulation results indicate that under the given condition, the proposed scheme can achieve a sum secrecy rate gain of more than 20 bps/Hz compared to the passive STAR-RIS-assisted cooperative jamming scheme, and an improvement of approximately 1 bps/Hz over the active STAR-RIS-assisted scheme without cooperative jamming, demonstrating that by virtue of the flexible signal regulation and active amplification capabilities enabled by the active STAR-RIS, the generated friendly jamming can interfere with eavesdroppers more effectively, thus improving the secure transmission quality of the system.Keywords:simultaneously transmitting and reflecting reconfigurable intelligent surface;cooperative jamming;secure transmission;beamforming;sum secrecy rate83|72|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:Semantic communication is one of the key approaches to achieving intelligent and efficient communication systems. However, complex and ever-changing wireless channel environments pose severe challenges to the robustness of semantic communication systems. To address this, this paper proposes a joint optimization method for channel parameter selection and synonym mapping tailored to Nakagami-m fading channels, specifically targeting the issue of semantic distortion caused by channel fading in multi-task semantic communication systems. First, a semantic mapping module is designed to dynamically represent the semantic features of text and images. Next, a joint loss function is formulated with the objective of maximizing semantic mutual information. Finally, an end-to-end task-adaptive algorithm is proposed to enhance the robustness of multi-task semantic transmission by jointly optimizing channel parameter selection and semantic mapping relationships. Experiments demonstrate that, compared to baseline methods, the proposed approach significantly reduces model complexity while effectively improving multi-task transmission performance. Specifically, the number of model parameters is reduced by 48.85%, and the classification task accuracy is increased by approximately 4% on average under low signal-to-noise ratio conditions. Furthermore, in the reconstruction task, the bilingual evaluation understudy (BLEU) score for text generation improved by approximately 0.1, and the peak signal-to-noise ratio (PSNR) for image reconstruction increased by approximately 4 dB.Keywords:semantic communication (SC);synonym mapping;channel parameters;task adaptation116|58|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:As an emerging low-cost, low-power green communication paradigm, backscatter communication has become one of the key candidate technologies for future Internet of Things (IoT) applications and has received increase attention in both industry and academia. However, the inherent double-depth fading characteristics of the backscatter channel can exacerbate signal distortion during transmission. Meanwhile, the terminal energy supply under the passive (battery free) communication mechanism completely depends on RF Energy Harvesting (EH), which further limites the transmission reliability and coverage of backscatter communication systems due to the inherent conflict between energy acquisition and communication performance. Although orthogonal space-time block code (OSTBC) can effectively mitigate the impact of double-depth fading on system reliability, the conventional design of OSTBC only focuses on the performance parameter of system reliability and fails to take into account the special mechanism of passive transmission of backscatter communication. To bridge this gap, this paper combines the passive communication mechanism of backscatter tags with the statistical characteristics of the backscatter channel to design a stacked rotating Alamouti code (SRAC) for quad-antenna backscatter tags, which can achieve the collaborative optimization of EH efficiency and system reliability. To the best of our knowledge, this is the first literature that incorporates EH performance into the design of backscatter space-time block code (STBC) for quad-antenna backscatter tags. The research results show that the proposed SRAC outperforms the conventional OSTBC in EH performance under both the linear EH model and the nonlinear EH model. This performance improvement benefits from the cyclic idle time slot in the algebraic structure of proposed SRAC, in which the backscatter tag antenna can harvest energy with maximum efficiency. In addition, this paper also derives the asymptotic closed-form expression of the symbol error rate (SER) of the proposed SRAC in the backscatter channel to specifically illustrate its performance. The results show that the proposed SRAC can fully utilize the inherent diversity gain of the backscatter channel when more than two receiving antennas are employed. Furthermore, with the same spectral efficiency, the proposed SRAC can also achieve 5~6 dB performance improvements in SER compared to the conventional OSTBC due to its full-rate structure. The above observations mean that the proposed SRAC is more friendly to battery free backscatter tags that require energy self-sustaining with respect to the convectional OSTBC. Numerical simulations are also conducted to verify the accuracy of our analysis.Keywords:backscatter communications;space-time block code;energy harvesting;diversity order;symbol error rate81|73|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:To address the multi-objective, high-dimensional, and strongly constrained mixed-integer combinatorial optimization problem encountered in non-regular subarray partitioning of large-scale phased array antennas, this paper proposes a non-regular subarray design method for large-scale phased arrays based on the maximum entropy criterion. By constructing a hierarchical optimization framework, the original complex problem is decomposed into two low-dimensional, relatively decoupled and hierarchically solvable subproblems, namely subarray-level topology optimization and module-level topology optimization, thereby effectively reducing the overall design complexity. In the subarray-level topology optimization stage, a small-scale phased array is adopted as the design carrier. Based on graph-theoretic principles, the array structure is modeled as an undirected graph, and the exact aperture coverage problem is abstracted as a matching problem in graph theory. On this basis, a constraint-guided random growth algorithm is introduced to achieve accurate subarray partitioning of the aperture. Furthermore, by exploiting the spatial distribution characteristics of subarray phase centers, information entropy is introduced as an evaluation metric to quantitatively assess different subarray partitioning schemes. Several subarray topologies with higher entropy values are then selected to construct a high-performance module library, which serves as a candidate set for subsequent module-level topology optimization. In the module-level topology optimization stage, modules are combined and arranged according to a predefined placement sequence, and the module combination scheme is treated as the optimization object. The quality of module placement is evaluated by calculating the information entropy of the phase-center distribution of the resulting large-scale phased array. Subsequently, the module placement sequence is encoded to form a chromosome structure, and a genetic algorithm is employed to search for the optimal solution of this discrete combinatorial optimization problem. Under the constraint of complete aperture coverage, a highly non-periodic array topology is achieved by maximizing the array entropy, thereby effectively suppressing the generation of grating lobes. Electromagnetic simulations are conducted on a practical array configuration operating at a center frequency of 30 GHz. The results demonstrate that the proposed method can achieve sidelobe levels below -10 dB within a beam scanning range of ± 60°. Compared with conventional non-periodic subarray layouts, the sidelobe level at large scanning angles can be reduced by up to 6.138 3 dB. Meanwhile, while maintaining array radiation performance, the number of transmit/receive (T/R) modules is reduced to 50% of that in conventional designs, significantly lowering the system cost and design complexity of large-scale phased arrays. By innovatively integrating the modular design concept with the maximum entropy optimization criterion, the proposed method provides a new technical approach for non-regular subarray design of large-scale phased arrays and exhibits strong engineering applicability and promising prospects for practical deployment.Keywords:large-scale phased arrays;non-regular subarray design;maximum entropy criterion;hierarchical optimization design;modular design;grating-lobe suppression;cost reduction86|59|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:As a core passive component in microwave and RF systems, the branch-line coupler is widely applied to key modules such as mixers, power distribution networks, and antenna feed networks. Traditional fabrication processes and low-order topologies typically suffer from inherent drawbacks, including narrow bandwidth and high-frequency loss, making them incapable of meeting the stringent requirements of modern communication systems for wide bandwidth and high efficiency. In contrast, high-order wideband topologies rely heavily on high-impedance transmission lines, which not only pose significant challenges to process implementation but also fail to satisfy the engineering demands of millimeter-wave systems for wide bandwidth and low loss. To address the aforementioned issues, this paper proposes an ultra-low-loss wideband millimeter-wave branch-line coupler based on micro-structure electrochemical fabrication (MEFAB), whose operating frequency band covers the 5G communication band of 24.5~41.3 GHz. In terms of process implementation, a three-dimensional air-dielectric copper-based electroforming molding technology is adopted to construct an enclosed micron-scale coaxial transmission line structure. With air serving as the primary dielectric and copper walls providing a fully shielded electromagnetic environment, the energy attenuation caused by dielectric loss and spatial radiation loss in the millimeter-wave band is fundamentally suppressed. From the perspective of topological structure, a third-order branch-line topology loaded with short-circuited matching stubs is adopted. It achieves electrical performance superior to that of conventional fourth-order topologies while avoiding the use of high-impedance transmission lines. Meanwhile, leveraging the superior stability of short-circuited structures in coaxial transmission lines, fabrication requirements are significantly reduced, and structural reliability is greatly enhanced. In the theoretical and design phase, odd-even mode analysis is first carried out on the topological circuit of the third-order branch-line coupler loaded with short-circuited stubs, and an initial physical model is established based on the theoretical topology. However, affected by structural discontinuities such as bends and T-junctions in the physical layout, a certain performance deviation exists between the established model and theoretical results. To eliminate such deviation, a chamfer compensation strategy is introduced for bent structures to weaken electromagnetic wave reflection at corners. Meanwhile, an equivalent circuit model is established for T-junctions, and the parasitic effects induced by discontinuities are effectively compensated through parameter extraction and iterative optimization of structural dimensions. The proposed compensation methods remarkably narrow the performance gap between theoretical simulation and practical physical structure.Keywords:micro-structure electrochemical fabrication (MEFAB);branch-line coupler;short-circuited stub matching network;T-junction optimization129|125|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:Following disasters, traditional communication infrastructure often suffers severe damage, leading to unstable network coverage and drastic fluctuations in link quality. Disaster emergency communication networks comprise various heterogeneous links, including residual ground base stations, drones, satellite links, and backpack equipment carried by rescue personnel. Multipath transmission technology improves network reliability and transmission efficiency by utilizing multiple heterogeneous links. However, the performance of multipath packet scheduling algorithms is highly dependent on real-time network conditions and changes with these conditions. The latency, bandwidth, packet loss rate, and jitter of links vary significantly across different disaster scenarios, making it difficult for a single fixed scheduling algorithm to maintain optimal performance in diverse and heterogeneous disaster network environments. Furthermore, disaster network scenarios are highly dynamic, with frequent changes in the status of links such as satellites, drones, and ground base stations, making it difficult for existing scheduling algorithms to respond quickly. Meanwhile, emergency communications require immediate deployment, making offline training based on large amounts of historical data impractical. To address these challenges, this paper proposes an online learning of adaptive optimal selection scheduling algorithm (OLAOS) based on the multipath quick UDP internet connections (MPQUIC) protocol. Rather than constructing a single universal scheduling algorithm, OLAOS dynamically selects the best among candidate scheduling algorithms based on current network conditions. It consists of three core modules: a scheduling algorithm candidate pool, an online learning selection module, and a network state awareness module. OLAOS treats multiple classic scheduling algorithms as a candidate set and models the multipath scheduling process as an online algorithm selection problem. Then, at the coarse-grained time slot level, OLAOS uses the contextual multi-armed bandit (CMAB) algorithm to select the optimal scheduling algorithm online based on real-time transmission status and historical performance feedback. At the fine-grained packet level, the selected scheduling algorithm is used to specifically execute packet scheduling. Furthermore, to address dynamic changes in links, OLAOS introduces a sliding window-based network awareness module. This module monitors fluctuations in the reward sequence to determine environmental changes and resets model parameters when significant fluctuations are detected, preventing historical models from failing after sudden environmental changes. Experimental results in multiple typical scenarios show that, compared to using a single, fixed scheduling algorithm, OLAOS can select the appropriate scheduling algorithm for the current scenario online based on the network status, achieving near-optimal overall performance in various scenarios. While maintaining lightweight implementation and easy deployment, OLAOS achieves relatively short completion time and high overall throughput, verifying its effectiveness and adaptability in heterogeneous and dynamic disaster emergency communication scenarios.Keywords:multipath transmission;disaster network;packet scheduling;QUIC;online learning;adaptive scheduling71|66|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:To address the collaborative requirements of high-security transmission and high quality of service (QoS) in dispersed computing network under harsh wireless communication environments, this paper proposes a secure communication and resource optimization scheme for intelligent reconfigurable surface (IRS)-assisted dispersed computing network. Firstly, due to the energy limitations of unmanned aerial vehicle (UAV) nodes, this paper studies a novel energy harvesting (EH) scheme. By dividing the IRS passive reflection array in geometric space, some reflection elements are used for information reflection, and some elements are used for EH, so as to realize the cooperation of information transmission and EH. Secondly, the secure sum rate maximization optimization model in IRS-assisted dispersed computing network is formulated. The model jointly optimizes multiple coupling variables such as user transmission power, IRS reflection element phase shift, EH constraint and communication QoS, while improve the overall system security performance and resource utilization efficiency. Since the formulated optimization problem is highly non-convex and the variables are strongly coupled, traditional optimization methods are difficult to directly obtain the global optimal solution. Furthermore, considering the characteristics of dispersed computing network, such as high user mobility, rapidly varying wireless channels, and uncertain environmental states, a robust deep reinforcement learning (DRL)-based dynamic resource optimization algorithm is designed to guarantee QoS in dynamic dispersed computing environments. Simulation results show that the performance of the IRS-assisted dispersed computing network scheme based on robust DRL proposed in this paper not only outperforms existing learning-based solutions but also achieves performance close to that of the exhaustive search method, verifying the effectiveness and superiority of the proposed scheme.Keywords:dispersed computing network;intelligent reflecting surface (IRS);energy harvesting (EH);secure sum rate;robust deep reinforcement learning (DRL)125|70|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:Tensor-based radiance field methods establish a mapping between inputs (3D spatial positions) and outputs (volume density and appearance features) via tensor regression. These methods rely on compact scene representations and significantly improve the efficiency of novel view synthesis while maintaining high-quality rendering results. However, existing approaches, whether based on conventional tensor decomposition or tensor train (TT) decomposition, are unable to fully exploit structural information in 3D scene space, thereby limiting the representation of deep-level features. To address this issue, we introduc tensor ring (TR) decomposition into the vector-matrix (VM) decomposition framework and proposed a vector-matrix tensor ring radiance fields (VMTR-RF) model for novel view synthesis. Unlike existing tensor radiance field methods, VMTR decomposition adopt a hierarchical modeling strategy: VM decomposition is first used to represent the scene as a combination of outer products of multiple vector and matrix factors, enabling an initial compact representation of the 3D scene. The vector-matrix factors are then reorganized into high-order tensors and further decomposed using TR decomposition, resulting in a tensor ring network composed of multiple core tensors, thereby enabling more effective capture of deep-level features in 3D scenes. Benefiting from the VMTR decomposition, VMTR-RF exhibit stronger modeling capability in volume density estimation and appearance feature learning. Finally, novel view synthesis is performed using volume rendering by combining the learned volume density and appearance features. Experimental results demonstrate that VMTR-RF outperformed existing state-of-the-art methods, particularly in detail preservation, enabling better reconstruction of sharp edges, complex structures, and natural textures, while achieving higher-quality novel view synthesis with a compact scene representation.Keywords:neural radiance fields;tensor network;novel view synthesis;compact representation;VMTR decomposition80|67|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:In AI-model network, edge-deployed small language models (SLMs) leverage token-level parallel collaboration to boost inference precision. However, limited by their own capabilities, relying solely on small models for collaboration can only achieve relatively limited performance gains.To overcome this, integrating large language model (LLM) to assist edge SLMs provides a robust solution for elevating the system’s performance ceiling. However, the resulting computational overhead is non-trivial, making cost-efficient high-performance inference a pressing challenge. To bridge this gap, we introduce DuGui, a consensus-based, LLM-guided mechanism. By maximizing the use of edge-based SLMs capabilities and resources, DuGui achieves superior inference performance without escalating cost. Specifically, the mechanism assesses predictive uncertainty by quantifying the degree of consensus between the token-vector distributions of the collaborating SLMs. When a divergence in consensus occurs, the LLM intervenes to provide guidance, thereby partially mitigating potential error accumulations and enhancing the inference performance of the SLMs. The mechanism adaptively executes distinct inference strategies based on the measured consensus: a high level of consensus indicates that the SLMs have converged on the current token, allowing for joint sampling without LLM intervention. Conversely, insufficient consensus triggers a routing request to the LLM to generate a guiding token sequence. By dynamically switching between collaborative SLMs inference and LLM-guided inference, DuGui leverages the advantages of edge-side collaboration while minimizing LLM invocation. Experimental results demonstrate that in the MMLU dataset for the qw1.5&glm combination, DuGui achieves 93% inference accuracy while incurring only 26.7% of the LLM’s output cost.This represents a 31% improvement in accuracy compared to the baseline SLMs. Consequently, the mechanism delivers inference performance nearly comparable to that of the LLM. Furthermore, compared to baseline methods, DuGui achieves a high-quality response proportion of 43.5%~63.9%, balancing cost-efficiency with high-performance execution.Keywords:AI-model network;large model guidance;small models consensus;model parallel inference;inference cost135|96|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:Alzheimer’s disease (AD) is a latent and irreversible progressive neurodegenerative disorder. Early precise identification is crucial for delaying disease progression and supporting clinical intervention. Structural magnetic resonance imaging (sMRI) can reveal anatomical abnormalities such as brain atrophy and gray matter degeneration, while Functional magnetic resonance imaging (fMRI) captures functional activities and dynamic interactions between brain regions. Both provide complementary information for AD diagnosis from structural and functional perspectives. Multimodal joint diagnosis for sMRI and fMRI faces three key challenges. First, 3D sMRI and 4D fMRI differ significantly in spatial resolution, temporal dimension, and signal distribution, making it difficult to construct a unified end-to-end cross-modal modeling framework. Second, although vision transformer-based sMRI feature extraction offers global modeling capabilities, the standard multi-head attention suffers from redundant heads, insufficient inter-head collaboration, and limited representation of local structural details, reducing sensitivity to AD-relevant regions such as the hippocampus and olfactory cortex. Third, most multimodal fusion methods rely on feature concatenation, unidirectional attention, or dense interaction strategies, which are insufficient to screen key regions and establish fine-grained bidirectional associations between structural and functional features in high-dimensional heterogeneous image data. To address these issues, this paper proposes BiSparseFusion, a cross-modal bidirectional sparse interaction fusion model. The sMRI branch employs a 3D Vision Transformer enhanced with a dynamic composable multi-head attention mechanism (DCMHA) and a multi-level feature fusion module (MFFM). DCMHA reduces redundant attention outputs by dynamically combining attention heads, and MFFM aggregates multi-level features to enhance local lesion details and global semantic representation. The fMRI branch uses SwiFT to directly model spatio-temporal dependencies of the original 4D fMRI, avoiding information loss caused by conventional region of interest- or connectivity-based methods. During cross-modal fusion, a bidirectional sparse cross-attention fusion module (BSCAF) suppresses redundant features within modalities and enables deep complementary interaction between sMRI structural and fMRI functional representations. The proposed method is validated on the ADNI dataset across three classification tasks: AD/NC, MCI/NC, and AD/MCI. BiSparseFusion achieves classification accuracies of 97.67%, 93.27%, and 96.72%, respectively, surpassing various single- and multi-modal comparison models. Visualization results indicate that the model effectively focuses on brain regions associated with AD pathology, including the hippocampus, olfactory cortex, and amygdala, forming a more discriminative fusion feature space. These results demonstrate the effectiveness of BiSparseFusion in multi-modal neuroimaging feature modeling, cross-modal fine-grained fusion, and AD auxiliary diagnosis.Keywords:Alzheimer’s disease;multimodal fusion;Structural magnetic resonance imaging;Functional magnetic resonance imaging;Dynamic Composable Multi-Head Attention;Bidirectional Sparse Cross-Attention Fusion94|89|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:Advanced persistent threats (APTs) have emerged as one of the most severe challenges to modern cybersecurity defense systems due to their extreme stealthiness, prolonged duration, and multi-stage nature. Although provenance-based analysis of host logs provides structured support for threat detection by correlating isolated system events into granular behavioral auditing paths, existing research still faces a core bottleneck: in complex system environments, attackers often disguise malicious activities as benign behavior through low-frequency operations, rendering traditional detection schemes based on signatures or static rules highly susceptible to failure when encountering zero-day attacks. To address these challenges, this paper proposes a hierarchical-aware graph masked autoencoder framework for APT detection. The primary innovation of this framework lies in the introduction of hierarchical topological knowledge to guide the masking process, fundamentally overcoming the limitations of blind random masking. Specifically, the model integrates three targeted strategies: global-aware masking (GAM), local-aware masking (LAM), and element-aware masking (EAM). GAM aims to preserve the macro-structural stability of the provenance graph; LAM focuses on characterizing neighborhood interaction logic between entities; and EAM addresses fine-grained entity attributes. This hierarchical design effectively filters out non-structural system noise during the pre-training phase while maximizing the retention of critical causal logic chains. Notably, the node-level consistency constraint models at an atomic scale, effectively circumventing the risk of signal dilution caused by global aggregation in traditional graph representation learning. This ensures that even under extremely imbalanced sample distributions, faint attack signals can still obtain sufficient gradient responses through the loss function, thereby mathematically guaranteeing the logical alignment between training objectives and point-wise anomaly detection tasks. During the detection phase, the framework employs an unsupervised anomaly detection algorithm to quantify node anomaly scores based on the embedding distributions of entity types, enabling the precise identification of malicious behaviors that disrupt local causal chains. Comprehensive evaluations were conducted on multiple authoritative public datasets, including StreamSpot, Unicorn Wget, and DARPA E3. Experimental results demonstrate that the proposed framework achieves an average precision of 98.49% and an F1-score of 98.97%. Compared to state-of-the-art baselines, our method exhibits superior robustness and recall in scenarios with extremely low attack base rates, effectively identifying subtle anomalous signals throughout the entire APT lifecycle.Keywords:advanced persistent threats;anomaly detection;graph masked autoencoder;graph neural network;self-supervised learning;System Provenance Graph84|79|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:PFP is an ultra-lightweight block cipher proposed by Huang Yuhua et al. in 2017. Owing to its simple structure, low implementation cost, and suitability for resource-constrained environments, it has attracted considerable attention in the area of lightweight cryptography. In the original design paper, the designers stated that at least 56 S-boxes are active over 15 rounds of PFP, and concluded that the maximum linear bias of 15-round PFP is upper-bounded by 2-57. According to the existing literature, the cryptanalysis of PFP has mainly focused on differential cryptanalysis, impossible differential cryptanalysis, and integral cryptanalysis, while a systematic, independent, and quantitative third-party evaluation of its resistance to linear cryptanalysis in the single-key setting is still lacking. To fill this gap, this paper presents the first comprehensive third-party linear cryptanalysis of PFP and obtains the best known distinguisher and key-recovery attack results in the single-key setting. Our analysis is based on an automatic search framework using mixed-integer linear programming (MILP). By taking into account the structural properties of PFP, including its Feistel architecture, PRESENT-type 4-bit S-box, and bit-permutation-based diffusion layer, we establish automatic search models for both linear characteristics and linear hulls tailored to this cipher. Specifically, the propagation behavior of linear masks through the basic components of PFP is encoded as a system of linear inequalities, and the objective function is designed to search for dominant linear characteristics with large correlation efficiently. With the aid of the Gurobi optimizer, we search not only for the best linear characteristic, but also for a large number of dominant linear characteristics under fixed input and output masks, so as to evaluate the corresponding linear hull more accurately. The results show that PFP admits a 26-round linear characteristic with correlation 2-31, which is currently the best distinguisher known for PFP in the single-key setting. Moreover, by fixing the input and output masks corresponding to this distinguisher, we further enumerate 670 dominant linear characteristics and estimate the expected linear potential (ELP) of the corresponding 26-round linear hull as 2-60.46. These findings indicate that PFP possesses stronger long-round linear distinguishers than previously reported. To support the validity of the theoretical estimation, we also conduct experimental verification for reduced-round cases. The experimental statistics are in good agreement with the theoretical analysis, which further confirms the reasonableness of the estimated expected linear potential and the effectiveness of the constructed linear approximations. Based on the 26-round linear distinguisher, we extend the attack by appending two rounds before the distinguisher and two rounds after it. Then, by combining the partial-sum technique with the key-bridging technique, we construct a single-key linear key-recovery attack on 30-round PFP. The resulting attack requires a data complexity of 263 known plaintexts and has an overall time complexity of approximately 278.92. In addition, under the 29-round setting, the time complexity of our work is reduced by 51.79% compared with that of the previous best approach. Therefore, the proposed method not only extends the best known single-key key-recovery attack on PFP by one more round, but also improves the attack efficiency when the number of attacked rounds is the same. A noteworthy point in our complexity evaluation is that, unlike some previous studies in which the exhaustive search phase was omitted from the total attack complexity, we uniformly include the exhaustive search cost determined by the attack advantage. In linear cryptanalysis, the effective amount of key information recovered in the guess-and-determine phase is determined by the advantage provided by the distinguisher, rather than merely by the nominal number of guessed key bits. Consequently, if the remaining exhaustive search cost is ignored, the reported complexity may underestimate the actual cost of a complete key-recovery attack. By incorporating this part into the total complexity, the present work provides a fairer and more objective basis for comparing different attack results.Combining the best linear distinguisher and key-recovery attacks obtained in this paper, we conclude that, in the single-key setting, the 34-round version of PFP retains only about 4 rounds of security margin, corresponding to approximately 11.8% of the full number of rounds. This result suggests that the resistance of PFP to linear cryptanalysis is weaker than previously understood and deserves further scrutiny. More importantly, the analysis presented in this paper provides a clearer quantitative understanding of the linear security of PFP and offers useful guidance for the security evaluation and design improvement of lightweight block ciphers with similar structures.Keywords:PFP;linear cryptanalysis;key recovery cryptanalysis;mixed-integer linear programming;partial-sum technique;key-bridging technique127|79|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:Frequency division power line communication (FD-PLC) technology effectively addresses the communication reliability and networking efficiency issues caused by the complex power line channel environment in local Industrial Internet of Things (IIoT) communication networks, owing to its flexible multi-band communication capability and efficient preamble-based contention access mechanism. However, existing hardware implementations of FD-PLC nodes suffer from high resource consumption and power dissipation, making them difficult to meet the practical requirements of low-cost and low-power IIoT communication nodes. To address this problem, this paper proposes a multi-frequency synchronization system based on dynamic band detection, which reconstructs the conventional multi-channel parallel detection architecture into an implementation combining real-time spectrum sensing and single-channel synchronization detection, thereby reducing hardware implementation complexity while maintaining nearly unchanged synchronization performance. The proposed system employs a pipelined Fast Fourier Transform (FFT) to achieve real-time power spectrum analysis. By performing a normalized division operation between the signal power spectrum and the background noise spectrum, the normalized accumulated power within each frequency band is calculated to realize dynamic blind detection of the signal band. Frame synchronization is then accomplished through an improved timing mechanism and a peak search algorithm. Simulation models are established under typical power line multipath channels as well as background noise, impulsive noise, and narrowband interference conditions. Measured power line noise is further introduced to validate the system performance, and comparisons are conducted with the conventional eight-channel parallel detection architecture for FD-PLC. The results show that, under high signal-to-noise ratio (SNR) conditions, the synchronization performance of the proposed system is basically consistent with that of the conventional scheme. Under low-SNR and various non-stationary noise conditions, the synchronization performance degradation is no more than 1 dB, the false alarm performance remains highly consistent with that of the conventional parallel architecture, and the proposed system exhibits good robustness under both narrowband interference and measured noise conditions. Complexity analysis shows that, with only about 3.43% additional bandwidth overhead, the proposed scheme reduces the numbers of real additions and real multiplications in the continuous monitoring stage by about 54% and 57%, respectively, and reduces those in the capture stage by about 23% and 28%, respectively. Furthermore, the proposed system is implemented on an Altera Cyclone Ⅳ E field-programmable gate array (FPGA) platform. Comprehensive results indicate that, compared with the conventional eight-channel parallel detection scheme, the proposed system reduces total logic element usage by about 72%, combinational logic functions by about 73%, dedicated logic registers by about 75%, embedded 9-bit multipliers by about 62%, and total power consumption by about 40%. The results demonstrate that the proposed system achieves an effective balance among synchronization performance, computational complexity, and hardware resource consumption, making it suitable for large-scale, low-cost, and low-power deployment of FD-PLC communication nodes in IIoT scenarios.Keywords:industrial internet of things;frequency division power line communication (FD-PLC);frame synchronization;multi-frequency detection;FPGA87|70|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:The missile-borne terminal guidance radar employs monopulse angle measurement to achieve precise angle tracking of a ship target during the tracking phase. Centroid corner reflector jamming exploits the inherent limitation of monopulse angle measurement — that it points to the energy centroid of the echo rather than the geometric center of the target — by deploying corner reflectors with high radar cross section (RCS) in the same range resolution cell as the ship, thereby decoying the tracking point and causing the missile to miss. In the range-Doppler image, the ship and the centroid corner reflector occupy the same range-Doppler cell, making identification and suppression of the corner reflector extremely difficult. The fundamental reason is that the two are closely spaced and aliased in the range-azimuth parameter domain, while polarization, spatial, and other multi-dimensional radar features have long been processed in isolation, leading to insufficient joint identification capability. Forward-looking imaging can integrate range-azimuth-polarization information and offers the potential to identify and localize the target and the jammer in the image domain. However, under forward-looking conditions, the platform motion direction is nearly aligned with the beam pointing direction, so the target Doppler frequency variation is minimal, rendering high-resolution methods that rely on the Doppler effect, such as synthetic aperture and Doppler beam sharpening, ineffective. Methods such as monopulse imaging, scanned beam deconvolution imaging, and array spectral estimation imaging face performance bottlenecks in close-range terminal centroid jamming scenarios, including pointing errors, scan dependence, and low signal-to-clutter ratio, respectively. Polarization, as a vector property of electromagnetic waves, contains rich information about the target’s geometric structure and scattering mechanisms. The polarization scattering characteristics of a ship and a corner reflector are inherently different. However, classical polarization processing methods such as polarization matching and polarization filtering are designed based on the enhancement or suppression of echo energy and fail to fully exploit the separability differences of signal components between targets in the polarization domain. This paper, grounded in the polarization modulation super-resolution theory, derives a polarization array echo model, establishes an adaptive modulation criterion, and proposes a parallel iterative optimization method based on polarization state screening. On this basis, by exploiting the differences in polarization scattering characteristics between the ship and the corner reflector, a corner reflector suppression method based on polarization principal components is proposed, thus forming a forward-looking imaging method against centroid corner reflector jamming. Simulation results show that at a signal-to-clutter ratio of 10 dB, for a combined target consisting of a trihedral and a dihedral corner reflector separated by 0.2 times the beamwidth, the proposed method achieves an angle measurement accuracy of 0.05 times the beamwidth. Under conditions of an 8 km missile-to-target range and a 20 dB signal-to-clutter ratio, the proposed forward-looking imaging method improves the ship-to-corner-reflector signal-to-interference ratio by 23.3 dB and 11.4 dB compared with the single-polarization method and the polarization-spatial joint method, respectively, before ship maneuvering, and by 14 dB and 2.4 dB, respectively, after ship maneuvering.Keywords:polarization modulation;forward-looking;super-resolution;centroid-based jamming;corner reflectors suppression82|92|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:In the distributed estimation framework based on wireless sensor networks, the conventional periodic time-triggered mechanism suffers from inherent drawbacks such as high communication frequency and massive redundant data exchange, which easily leads to excessive consumption of network communication resources and further degrades the performance of distributed estimation. To address this issue, this paper proposes a stochastic event-triggered communication mechanism and investigates the event-triggered distributed nonlinear estimation method for wireless sensor networks. This paper focuses on the event-triggered distributed nonlinear filtering over wireless sensor networks by introducing a stochastic event-triggered communication scheme. Given the pronounced nonlinear characteristics of the system’s observation model, traditional linear estimation methods fail to match the state evolution rules of nonlinear systems and cannot guarantee satisfactory estimation performance. Thus, within the framework of unscented kalman filtering (UKF), this paper adopts the unscented transformation to accurately approximate the nonlinear state propagation process, and develops a UKF-based local estimator tailored for nonlinear observation models to effectively improve local estimation performance. To reduce excessive estimation interactions caused by continuous and frequent information interaction among neighboring nodes, a local estimation-based stochastic event-triggered communication strategy is constructed according to the characteristics of local state estimation errors. Meanwhile, a recursive event-triggered local estimator is designed based on Bayesian estimation theory. By further employing the covariance intersection fusion technique, event-triggered local estimates are fused in real time to derive an event-triggered distributed estimator, which guarantees the consistency of distributed estimation under incomplete information exchange among neighboring nodes. Furthermore, the convergence of the proposed event-triggered distributed estimator is analyzed under the global observability of the network, and a sufficient condition is established to ensure the boundedness of the distributed covariance. Finally, simulation experiments are conducted using a typical maneuvering target tracking scenario to verify that the proposed algorithm can significantly reduce the network communication load while ensuring estimation accuracy.Keywords:stochastic event-triggered communication;unscented Kalman filtering (UKF);covariance intersection fusion;nonlinear estimation;wireless sensor networks123|84|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:To tackle the challenges of obtaining precise radio environment map due to wide-ranging monitoring and complex terrain in electromagnetic environment surveillance, an efficient federated learning algorithm for multiple unmanned aerial vehicles (UAVs) collaborative radio environment map generation that integrates generative adversarial network (GAN) and physical-informed neural network (PINN) was considered. The proposed algorithm utilizes a framework named terminal device-edge computing node-UAV GAN-PINN federated learning. Firstly, the algorithm integrates GAN and PINN into Federated Learning to enhance the convergence of radio environment map generation models. This integration boosts the models’ generalization capabilities and ensures their physical consistency. Next, the algorithm takes into account various factors such as UAV bandwidth resource allocation, communication channels, model aggregation, flight trajectories, and edge node computational resources. It then formulates an optimization problem with the objective of maximizing model precision while considering constraints related to latency, energy consumption, and bandwidth. The algorithm solves this problem using multi-agent deep deterministic policy gradient algorithm, enabling UAV to make effective decisions in dynamic environments. Simulation results indicate that the proposed method achieves better performance in terms of both model accuracy and convergence speed. Additionally, it effectively optimizes UAV latency and energy consumption.Keywords:radio environment map;multi-agent deep deterministic policy gradient;federated learning;generative adversarial networks;physical-informed neural networks;unmanned aerial vehicle103|75|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:With the development of 5G wireless communication, the Internet of Things, high-speed interconnection, and artificial intelligence technologies, radio-frequency electronic devices have imposed higher requirements on the high-frequency transmission capability, flexible integration characteristics, and process compatibility of interconnect materials. Conventional metallic conductors suffer from problems such as high elastic modulus, relatively large mass, and additional losses induced by surface roughness in miniaturized, high-frequency, and flexible electronic applications, making it difficult for them to meet the development requirements of next-generation integrated systems. Two-dimensional layered titanium carbide (Ti₃C₂Tₓ, MXene) exhibits promising application potential in radio-frequency transmission lines and high-frequency interconnect devices owing to its relatively high electrical conductivity, lightweight nature, flexibility, and solution processability. However, studies on the effects of the film-forming process, surface morphology, and roughness characteristics of MXene films on radio-frequency transmission performance remain relatively limited. In this study, MXene thin-film coplanar waveguide radio-frequency transmission lines were investigated. The conductor surface roughness was regulated by controlling the preparation process, and its influence on radio-frequency transmission performance was studied. MXene films with thicknesses of 1.4 ± 0.2 μm,2.6 ± 0.3 μm, and3.6 ± 0.2 μm were prepared using doctor-blading and drop-casting methods. Atomic force microscopy, scanning electron microscopy, vector network analyzer measurements, and electromagnetic simulation were combined to analyze surface roughness, laser-cut edge morphology, insertion loss, and attenuation rate per unit length. The results show that different film-forming processes significantly affect the stacking state of MXene nanosheets, film uniformity, and surface roughness. The drop-casting process mainly relies on capillary-force-driven random assembly of nanosheets, which tends to form a rough surface with large height fluctuations. In contrast, the doctor-blading process induces the in-plane alignment of nanosheets through shear action, facilitating the formation of a flatter and denser film structure. For all three thicknesses, the surface roughness of the doctor-bladed films was only approximately 30%~40% of that of the drop-cast films. Radio-frequency test results show that lower surface roughness helps reduce the insertion loss and attenuation rate of MXene transmission lines, and this advantage becomes more pronounced at smaller film thicknesses and higher operating frequencies. At a thickness of 1.4 μm and a frequency of 18 GHz, the doctor-bladed sample exhibited an insertion loss approximately 2.7 dB lower than that of the drop-cast sample, with transmission performance improved by approximately 36.5% and attenuation rate reduced by approximately 1.7 dB/cm. Further results indicate that the roughness of the laser-cut edge also affects transmission performance, and a laser power of 25W with a cutting speed of 150 mm/s produced a more regular edge morphology and better transmission performance. Finally, by combining the roughness-gradient conductivity model, the influence of surface roughness on the transmission coefficient was revealed. This study provides experimental and theoretical bases for the process optimization and device design of radio-frequency transmission lines in electronic devices based on non-metallic, novel layered conductive materials.Keywords:MXene;Radio-Frequency transmission;surface roughness;thin-film fabrication;Transmission coefficient79|64|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:Low-light hazy scenarios suffer from composite image degradation stemming from insufficient illumination, haze occlusion and local glow scattering. Such intricate degradation severely compromises the visual perception performance of intelligent equipment including vehicles, unmanned aerial vehicles (UAVs) and robots, constraining their applicable scenarios and limiting their practical deployment under low-light hazy conditions. Therefore, restoring high-quality clear images from extreme degradation caused by nighttime dim light, haze veiling and local glows, as well as improving the all-weather safety and reliability of autonomous driving, UAV inspection and security monitoring systems, is a critical and pressing research challenge. To tackle this problem, this paper first investigates the physical perturbation mechanisms of illumination and haze in imaging, and explores the latent correlations between depth information and degradation disturbances induced by illumination and haze. On this basis, a low-light dehazing enhancement network guided by deep information-assisted gaussian mixture physics (DGP-Net) is proposed. This network effectively mitigates the composite degradation triggered by multiple light sources, haze interference and noise, thereby achieving robust enhancement for low-light hazy images. Specifically, this work makes three key technical contributions as follows. First, leveraging the inherent features of degraded images and depth maps, a gaussian hybrid-physics perturbation system (GHPPS) based image re-degradation paradigm embedded with depth information is constructed. By applying Gaussian blurring, Gaussian noise injection and polarized light disturbance to raw degraded images, a set of randomly independent perturbed images conforming to the Gaussian hybrid physical model is generated. Such re-degraded samples provide sufficient data augmentation for network training. Second, a depth-guided blended physical prior mechanism (DBPM) is elaborately designed. This mechanism directs the network to capture shared latent structural characteristics between original images and re-degraded image groups. With the auxiliary depth information, the network can precisely estimate the illumination component of degraded images. Moreover, a lightweight perturbation cross-attention (PCA) module is developed to facilitate the efficient learning of composite degradation features from both original and re-degraded images, which further strengthens the network’s perception ability of complex degradation patterns. Third, a spatial physics-guided frequency dehazing block (SPGFD) is established. This module integrates degradation feature representations from both spatial and frequency domains, overcoming the inherent drawbacks of single-domain enhancement methods that solely rely on either spatial or frequency characteristics. Equipped with this module, the network achieves favorable haze removal and glow suppression capabilities under low-light conditions, enabling accurate structural reconstruction and global illumination restoration for hazy scenes. Extensive experiments conducted on the low-light hazy NHR dataset demonstrate that the proposed DGP-Net outperforms the state-of-the-art (SOTA) CoA method. Specifically, the proposed method achieves a 11.3% improvement in PSNR and a 5.8% improvement in SSIM, yielding superior quantitative metrics and visually pleasing enhancement results.Keywords:Low-Light Hazy Image Enhancement;Gaussian Mixture Physical Prior Modeling;Perturbation Cross-Attention;Joint Space-Frequency Domain89|82|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
PAPERS
- Abstract:With characteristics of instantaneity, openness, and widespread dissemination, social networks facilitate information exchange but have simultaneously become a hotbed for the proliferation of rumors. Integrating heterogeneous information forms such as text and images, multimodal rumors exhibit greater deceptiveness and provocativeness, thereby significantly complicating network governance. Consequently, achieving efficient and accurate multimodal rumor detection has emerged as a critical mission for maintaining a clean cyberspace and ensuring public safety. In recent years, the research community has proposed diverse detection methodologies tailored to the data characteristics of multimodal rumors, leading to an increasingly diversified technical landscape. To systematically review current research, this paper constructs a multimodal rumor detection taxonomy from the perspective of information composition and sources, encompassing four dimensions: content, external knowledge, social context, and external environment. Unlike existing reviews with a singular emphasis on the content dimension, this paper, through its four-dimensional analytical framework, reveals a profound paradigm shift in rumor detection from localized static content verification to an integrated approach that incorporates dynamic social evolution and external environmental cross-validation. Within this framework, it analyzes the core mechanisms, evolutionary trajectories, and representative models of four methodological categories: visual-textual interaction, external knowledge enhancement, social context information, and external environment perception, summarizing their technical advantages and limitations. Additionally, this paper summarizes commonly used datasets and evaluation metrics, and comparatively analyzes the performance disparities of various models from the three dimensions of longitudinal technical evolution trajectories, cross-category statistical trends, and model performance under specific detection scenarios. Finally, this paper discusses five critical challenges: bottlenecks in modality heterogeneity and feature alignment, risks of external knowledge dependence and large language model (LLM) hallucinations, complexities of dynamic social and environmental modeling, cross-domain biases, and the authentication of highly realistic artificial intelligence generated content (AIGC) rumors. Furthermore, it outlines future research directions.Keywords:social networks;multi-modal rumor detection;multi-modal fusion;knowledge enhancement;social context;external environment perception179|146|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
- Abstract:As 6G communication evolves toward higher frequency bands, larger bandwidths, intelligent convergence, and dynamic reconfigurability, traditional fixed-position antenna systems are increasingly constrained by their static deployment and limited parameter-tuning capability. These limitations hinder their ability to cope with the variability of complex electromagnetic environments, the randomness of user mobility, and the inherent trade-off between energy efficiency and reliability. Movable antenna (MA) technology introduces the capability of physically adjusting antenna positions, thereby providing a promising technical pathway to overcome the static spatial degrees-of-freedom constraints of traditional fixed-position antenna systems. However, MA technology should not be regarded as a simple extension of conventional antenna architectures, since its performance depends on the coupled interaction among electromagnetic responses, information transmission, and position control. Therefore, it is essential to establish a unified modeling and collaborative optimization framework that captures the holistic structure, coupled interactions, and dynamic evolution of MA-enabled wireless systems. From a systems-theoretic perspective, this paper positions MA technology as the core enabler and proposes a cross-domain theoretical framework centered on the integration of electromagnetics, information theory, and cybernetics, aiming to establish a new paradigm for collaborative design and dynamic optimization in next-generation wireless communication systems. To this end, this paper first reviews the fundamentals of traditional fixed-position antenna systems, MA technology, and the roles of electromagnetics, information theory, and cybernetics in wireless system design. The subsequent discussion analyzes the limitations of single-theory frameworks in characterizing the dynamic optimization problem of MA systems, summarizes recent progress in tri-domain integration, and elaborates on the intrinsic logic, key challenges, and implementation pathways of the cross-domain integration of electromagnetics, information theory, and cybernetics. The essence of this framework lies in using electromagnetics to characterize physical channel responses, information theory to define system performance limits, and cybernetics to enable real-time dynamic regulation. These three theoretical domains are not isolated or parallel tools. Instead, they are deeply coupled and coordinated in a closed-loop manner under system-level objectives. Such integration provides a fundamental pathway for wireless communication to evolve from an experience-driven paradigm toward a model-driven paradigm. Furthermore, this paper systematically discusses recent advances in position-dependent channel modeling, system capacity optimization, intelligent position control, mechanical movement constraints, hardware implementation, and cross-layer collaborative design. Finally, this paper identifies potential research directions for MA technology empowered by tri-domain integration under the guidance of systems theory, including the coupling laws between MA dynamics and electromagnetic responses, intelligent MA control methods for communication performance enhancement, joint optimization of mechanical movement delay and transmission latency, and system-level MA design for practical deployment. The proposed perspective is expected to provide theoretical support and methodological guidance for the evolution of 6G communication toward sensing-native communication, intelligent networks, and reconfigurable environments.Keywords:electromagnetics;information theory;Cybernetics;movable antenna;closed-loop collaboration;cross-domain integration;intelligent regulation200|139|0<HTML><L-PDF> <Meta-XML>Updated:2026-10-16
SURVEYS AND REVIEWS








