Search Results (651)

Most electric vehicles use 400 V batteries, while some companies are moving to 800 V to reduce current in electric drives. More cars are expected to adopt 800 V at the DC terminals of the batteries, but 400 V will remain common for the duration of this transition, so future off-board chargers must support a wide voltage output range. Silicon carbide switches are used to keep the power–electronics interface compact and scalable. The AC/DC stage of a modular silicon carbide-based interface is designed using a T-type active front end and a dual active bridge. The T-type front end is optimized with a genetic algorithm. The resulting model is used to tune the inner current and outer voltage controllers. Bode analysis shows an inner current loop bandwidth of 4.25 kHz with a phase margin of ${53}^{\circ }$ and a gain margin of 30 dB. The outer voltage loop reaches 50 Hz with a phase margin of ${108}^{\circ }$ and a gain margin of 33 dB. The controller is implemented on a dSPACE MicroLabBox. Tests show peak efficiency of 98.5% in G2V mode and 98.3% V2G mode. THD stays under 5% above 4 kW and reaches 3% at peak power. Full article

Unmanned Aerial Vehicle (UAV) platforms enable flexible and cost-effective vehicle detection for intelligent transportation systems, yet small-scale vehicles in complex aerial scenes pose substantial challenges from extreme scale variations, environmental interference, and single-sensor limitations. We present AMSRDet (Adaptive Multi-Scale Remote Sensing Detector), an adaptive multi-scale detection network fusing infrared (IR) and visible (RGB) modalities for robust UAV-based vehicle detection. Our framework comprises four novel components: (1) a MobileMamba-based dual-stream encoder extracting complementary features via Selective State-Space 2D (SS2D) blocks with linear complexity $O\left(HWC\right)$, achieving 2.1× efficiency improvement over standard Transformers; (2) a Cross-Modal Global Fusion (CMGF) module capturing global dependencies through spatial-channel attention while suppressing modality-specific noise via adaptive gating; (3) a Scale-Coordinate Attention Fusion (SCAF) module integrating multi-scale features via coordinate attention and learned scale-aware weighting, improving small object detection by 2.5 percentage points; and (4) a Separable Dynamic Decoder generating scale-adaptive predictions through content-aware dynamic convolution, reducing computational cost by 48.9% compared to standard DETR decoders. On the DroneVehicle dataset, AMSRDet achieves 45.8% mAP@0.5:0.95 (81.2% mAP@0.5) at 68.3 Frames Per Second (FPS) with 28.6 million (M) parameters and 47.2 Giga Floating Point Operations (GFLOPs), outperforming twenty state-of-the-art detectors including YOLOv12 (+0.7% mAP), DEIM (+0.8% mAP), and Mamba-YOLO (+1.5% mAP). Cross-dataset evaluation on Camera-vehicle yields 52.3% mAP without fine-tuning, demonstrating strong generalization across viewpoints and scenarios. Full article

Industrial gateways serve as critical data aggregation points within the Industrial Internet of Things (IIoT), enabling seamless data interoperability that empowers enterprises to extract value from equipment data more efficiently. However, their role exposes a fundamental trade-off between computational efficiency and prediction accuracy—a contradiction yet to be fully resolved by existing approaches. The rapid proliferation of IoT devices has led to a corresponding surge in network traffic, posing significant challenges for traffic forecasting methods, while deep learning models like Transformers and GNNs demonstrate high accuracy in traffic prediction, their substantial computational and memory demands hinder effective deployment on resource-constrained industrial gateways, while simple linear models offer relative simplicity, they struggle to effectively capture the complex characteristics of IIoT traffic—which often exhibits high nonlinearity, significant burstiness, and a wide distribution of time scales. The inherent time-varying nature of traffic data further complicates achieving high prediction accuracy. To address these interrelated challenges, we propose the lightweight and theoretically grounded DOA-MSDI-CrossLinear framework, redefining traffic forecasting as a hierarchical decomposition–interaction problem. Unlike existing approaches that simply combine components, we recognize that industrial traffic inherently exhibits scale-dependent temporal correlations requiring explicit decomposition prior to interaction modeling. The Multi-Scale Decomposable Mixing (MDM) module implements this concept through adaptive sequence decomposition, while the Dual Dependency Interaction (DDI) module simultaneously captures dependencies across time and channels. Ultimately, decomposed patterns are fed into an enhanced CrossLinear model to predict flow values for specific future time periods. The Dream Optimization Algorithm (DOA) provides bio-inspired hyperparameter tuning that balances exploration and exploitation—particularly suited for the non-convex optimization scenarios typical in industrial forecasting tasks. Extensive experiments on real industrial IoT datasets thoroughly validate the effectiveness of this approach. Full article

During viral infection, NLR family CARD domain-containing protein 5 (NLRC5) participates in innate immunity through multiple mechanisms. These include regulating type I interferon and related immune factor expression, as well as modulating immune cell functions, such as cytotoxic T lymphocytes (CTLs) and macrophages, thereby promoting antiviral defence and maintaining immune homeostasis. Our study demonstrates that (1) Enterovirus 71 (EV71) infection upregulates NLRC5 expression through the RIG-I-IRF3-mediated IFN-β pathway, which in turn promotes MHC-I molecule expression and (2) NLRC5 suppresses EV71 replication and simultaneously restrains excessive inflammatory responses by fine-tuning IFN-β production through a negative feedback loop. This loop operates via two distinct mechanisms, namely, direct downregulation of key IFN-β pathway mediators (e.g., RIG-I and IRF3) and binding to the 5′UTR of the EV71 genome to inhibit viral replication, thereby indirectly dampening the IFN-β signal. Furthermore, we show that EV71 activates the NLRC5-dependent MHC-I response in an IFN-β-dependent manner. Collectively, these results elucidate the dual role of NLRC5 during EV71 infection, offering novel insights into viral pathogenesis and highlighting potential targets for antiviral drug development. Full article

A central challenge in reconfigurable photonics based on quasi bound states in the continuum (quasi-BICs) is to move beyond binary switching toward multistate and polarization-aware programmability. Here we propose a dual-phase-change material (PCM) metasurface that enables four-state nonvolatile switching and polarization-divergent dispersion rewriting within a single unit cell. Two independently switchable PCM layers provide four addressable configurations (0-0, 0-1, 1-0, 1-1) at a fixed geometry, allowing the resonance landscape to be reprogrammed through complex-index rewriting without structural modification. Angle-resolved transmission maps reveal fundamentally different evolution pathways for orthogonal polarizations. For p polarization, the quasi-BIC exhibits strong state sensitivity with dispersion reshaping and multi-branch features near normal incidence; the resonance red-shifts from ~1331 nm to ~1355 nm while the quality factor decreases from ~6.7 × 10⁴ to ~4.0 × 10⁴. In contrast, for s polarization, a single weakly dispersive branch translates coherently across states, producing a much larger shift from ~1635 nm to ~1790 nm while the quality factor increases from ~9.0 × 10³ to ~1.8 × 10⁴. The opposite quality-factor trajectories, together with the polarization-contrasting tuning ranges, demonstrate that dual-PCM programming reconfigures polarization-selective radiative coupling rather than imposing a uniform resonance shift. This compact two-bit metasurface platform provides multistate, high-Q control with active dispersion engineering, enabling polarization-multiplexed reconfigurable filters, state-addressable sensors, and other programmable photonic devices. Full article

Achieving a balance between rapid adaptation and robustness is a critical yet challenging objective in the design of industrial control systems. Model Reference Adaptive Control (MRAC) is a standard approach for managing system uncertainties; however, it suffers from a fundamental trade-off between adaptation speed and robustness. The high adaptation gains required for fast tracking often lead to parameter bursting or instability in the presence of noise. To resolve this issue, this paper proposes a new Dual-Loop Optimal Feedback Control (OFC) architecture applied to a Brushless DC (BLDC) motor drive. Unlike conventional methods that rely solely on tuning the adaptive mechanism, the proposed architecture introduces a parallel compensation loop designed to decouple disturbance rejection from reference tracking. This structure utilizes a Genetic Algorithm (GA) as an offline optimization engine to identify the Optimal Compensator gains that balance transient recovery with steady-state stability. Experimental validation demonstrates that the proposed Dual-Loop OFC architecture significantly outperforms traditional approaches. Specifically, it achieves an 88.99% reduction in overshoot and a 13.8% reduction in settling time compared to Conventional MRAC (CMRAC). Furthermore, it exhibits an 86.7% faster rise time compared to Self-Tuning Fuzzy PID (STFPID). These results confirm that the proposed Dual-Loop structure effectively mitigates the classic adaptability–robustness trade-off, offering a stable and high-performance solution for industrial actuators under varying operating conditions. Full article

Semantic segmentation of high-resolution remote sensing imagery is a critical technology for the intelligent interpretation of sensor data, supporting automated environmental monitoring and urban sensing systems. However, processing data from dense urban scenarios remains challenging due to sensor signal occlusions (e.g., shadows) and the complexity of parsing multi-scale targets from optical sensors. Existing approaches often exhibit a trade-off between the accuracy of global semantic modeling and the precision of complex boundary recognition. While the Segment Anything Model (SAM) offers powerful zero-shot structural priors, its direct application to remote sensing is hindered by domain gaps and the lack of inherent semantic categorization. To address these limitations, we propose a dual-branch cooperative network, PriorSAM-DBNet. The main branch employs a Densely Connected Swin (DC-Swin) Transformer to capture cross-scale global features via a hierarchical shifted window attention mechanism. The auxiliary branch leverages SAM’s zero-shot capability to exploit structural universality, generating object-boundary masks as robust signal priors while bypassing semantic domain shifts. Crucially, we introduce a parameter-efficient Scaled Subsampling Projection (SSP) module that employs a weight-sharing mechanism to align cross-modal features, freezing the massive SAM backbone to ensure computational viability for practical sensor applications. Furthermore, a novel Attentive Cross-Modal Fusion (ACMF) module is designed to dynamically resolve semantic ambiguities by calibrating the global context with local structural priors. Extensive experiments on the ISPRS Vaihingen, Potsdam, and LoveDA-Urban datasets demonstrate that PriorSAM-DBNet outperforms state-of-the-art approaches. By fine-tuning only 0.91 million parameters in the auxiliary branch, our method achieves mIoU scores of 82.50%, 85.59%, and 53.36%, respectively. The proposed framework offers a scalable, high-precision solution for remote sensing semantic segmentation, particularly effective for disaster emergency response where rapid feature recognition from sensor streams is paramount. Full article

Hardware-in-the-loop (HIL) simulation is an essential tool for rapid and cost-effective development and validation of power-electronic systems. The primary objective of this work is to validate and fine-tune a PLECS-based HIL model of a single dual active bridge (DAB) DC-DC converter, thereby laying the foundation for building more complex models (e.g., multiple converters connected in series or parallel). To this end, the converter is experimentally characterized, and the HIL model is validated across a wide range of operating conditions by varying the PWM phase-shift angle, voltage gain, switching frequency, and leakage inductance. Power transfer and efficiency are analyzed to quantify the influence of these parameters on converter performance. These experimental trends provide insight into the optimal modulation range and the dominant loss mechanisms of the DAB under single phase shift (SPS) control. A detailed comparison between HIL simulations and hardware measurements, based on transferred power and efficiency, shows close agreement across all the tested operating points. These results confirm the accuracy and robustness of the proposed HIL model, demonstrate the suitability of the PLECS platform for DAB development and control validation, and support its use as a scalable basis for more complex multi-converter studies, reducing design time and prototyping risk. Full article

Fibrin clots with strain-hardening characteristics exhibit pronounced material nonlinearity and acoustic dispersion under ultrasound, leading to waveform distortion and shock formation during finite-amplitude wave propagation. However, peak-shock stress is limited by viscoelastic dissipation and dispersion, constraining the efficiency of ultrasound in applications such as thrombus ablation. To overcome this limitation, a shock wave amplification method using designed multi-wave-packet sequences is proposed. Based on a power-law model from quasi-static compression tests, shock generation under a single sinusoidal pulse was first simulated. The dual-wave-packet chasing strategy was then developed, in which the amplitude, frequency, and time delay of the second packet were tuned to achieve effective superposition with the precursor. The waveform superposition factor (WSF) was introduced for quantitative evaluation. Numerical results demonstrate that this strategy can significantly increase the peak-shock-wave stress, with a maximum gain of 22.7%. Parametric analysis further identified amplitude as the dominant factor influencing wavefront steepness and amplification effectiveness. This study provides a novel method and theoretical support for developing efficient and controllable ultrasonic sequences for thrombolysis. Full article

High-quality code representations are fundamental to code intelligence. Achieving such representations with parameter-efficient fine-tuning (PEFT) remains a key challenge. While code pre-trained models (CodePTMs) offer a robust foundation for general-purpose embeddings, current PEFT approaches face two main obstacles when adapting them: (i) they fail to adequately capture the deep structural characteristics of programs, and (ii) they are limited by the model’s finite internal parameters, restricting their ability to overcome inherent knowledge bottlenecks. To address these challenges, we introduce a parameter-efficient code representation learning framework that combines retrieval augmentation with structure-aware priors. Our framework features three complementary, lightweight modules: first, a structure–semantic dual-channel retrieval mechanism that infuses high-quality external code knowledge as non-parametric memory to alleviate the knowledge bottleneck; second, a graph relative bias module that strengthens the attention mechanism’s capacity to model structural relationships within programs; and third, a span-discriminative contrastive objective that sharpens the distinctiveness and boundary clarity of span-level representations. Extensive experiments on three benchmarks spanning six programming languages show that our method consistently outperforms state-of-the-art parameter-efficient baselines. Notably, on structure-sensitive tasks using the PLBART backbone, RS-Rep surpasses full fine-tuning, delivering a 22.1% improvement in Exact Match for code generation and a 4.4% increase in BLEU scores for code refinement, all while utilizing only about 5% of the trainable parameters. Full article

We experimentally and theoretically demonstrate coexistence of three different wavelength-multiplexed bound dual-frequency pulses in an all-fiber mode-locked fiber laser, effectively achieved by exploiting polarization-dependent loss effects and two uneven gain peaks of Er-doped fiber. With the single wall carbon-nanotube-based intensity modulation, wavelength-multiplexed dual-frequency pulses located at 1531.1 nm and 1556.6 nm are obtained. Changing the polarization rotation angles in the fiber cavity, one of the two asynchronous pulses evolves into a bound state of a doublet, in which the center wavelength of the bound solitons is centered at ~1530 nm or ~1556 nm. The relative phase between the two bound solitons or modulation depth of bound solitons can be switched by a polarization controller. A simulation method based on coupled Ginzburg–Landau equations is provided to characterize the laser physics and understand the mechanism behind the dynamics of tuning between different bound dual-frequency pulses. The proposed fiber laser will provide a potential way to understand multiple soliton dynamics and implementation in optical frequency combs generation. Full article

This study utilizes COMSOL Multiphysics (version 6.0) to design a planar ultra-broadband optical absorber with a multilayer configuration. The proposed structure consists of seven stacked layers arranged from bottom to top: W (h1, acting as a reflective substrate and transmission blocker), WSe₂ (h2), SiO₂ (h3), Ni (h4), SiO₂ (h5), Mo (h6), and SiO₂ (h7). One key finding of this study is that, when all other layer thicknesses are fixed, variations in the Mo layer thickness systematically induce a redshift in both the short- and long-wavelength cutoff edges. Notably, the long-wavelength cutoff exhibits a larger shift than the short-wavelength edge, resulting in an increased absorption bandwidth where absorptivity remains above 0.900. The second contribution is the demonstration that this planar structure can be readily engineered to achieve ultra-broadband absorption, spanning from the near-ultraviolet and visible region (360 nm) to the mid-infrared (6300 nm). An important characteristic of the proposed design is that the thickness of the h7 SiO₂ layer influences the cutoff wavelength at the short-wavelength edge, while the thickness of the h6 Mo layer governs the cutoff position at the long-wavelength edge. This dual modulation capability allows the proposed optical absorber to flexibly tune both the spectral range and the bandwidth in which absorptivity exceeds 0.900, thereby enabling the realization of a wavelength- and bandwidth-tunable optical absorber. Full article

Background/Objectives: Cognitive decline and compromised attention control serve as early indicators of neurodysfunction that manifest as broader psychopathological symptoms, yet conventional mental health assessment relies predominantly on subjective self-report measures lacking objectivity and temporal granularity. We propose Neurosense, an AI-driven brain health assessment framework using electroencephalography (EEG) to non-invasively capture neural dynamics. Methods: Our Dual-path Spatio-Temporal Adaptive Gated Encoder (D-STAGE) architecture processes temporal and spatial EEG features in parallel through Transformer-based and graph convolutional pathways, integrating them via adaptive gating mechanisms. We introduce a two-stage paradigm: first training on cognitive task EEG for reaction time prediction to acquire cognitive performance-related representations, then featuring parameter-efficient adapter-based transfer learning to estimate p-factor—a transdiagnostic psychopathology dimension. The adapter-based transfer achieves competitive performance using only 1.7% of parameters required for full fine-tuning. Results: The model achieves effective reaction time prediction from EEG signals. Transfer learning from cognitive tasks to mental health assessment demonstrates that cognitive efficiency representations can be adapted for p-factor prediction, outperforming direct training approaches while maintaining parameter efficiency. Conclusions: The Neurosense framework reveals hierarchical relationships between neural dynamics, cognitive efficiency, and mental health dimensions, establishing foundations for a promising computational framework for mental health assessment applications. Full article

Text-to-image (T2I) generation, a core component of generative artificial intelligence(AI), is increasingly important for creative industries and human–computer interaction. Despite impressive progress in realism and diversity, diffusion models still exhibit critical security blind spots particularly in the Transformer key-value mapping mechanism that underpins cross-modal alignment. Existing backdoor attacks often rely on large-scale data poisoning or extensive fine-tuning, leading to low efficiency and limited stealth. To address these challenges, we propose two efficient backdoor attack methods AttnBackdoor and SemBackdoor grounded in the Transformer’s key-value storage principle. AttnBackdoor injects precise mappings between trigger prompts and target instances by fine-tuning the key-value projection matrices in U-Net cross-attention layers (≈5% of parameters). SemBackdoor establishes semantic-level mappings by editing the text encoder’s MLP projection matrix (≈0.3% of parameters). Both approaches achieve high attack success rates (>90%), with SemBackdoor reaching 98.6% and AttnBackdoor 97.2%. They also reduce parameter updates and training time by 1–2 orders of magnitude compared to prior work while preserving benign generation quality. Our findings reveal dual vulnerabilities at visual and semantic levels and provide a foundation for developing next generation defenses for secure generative AI. Full article

This work presents a notch filter architecture based on a dual-output second-generation voltage conveyor, designed with a current-mode approach. The proposed topology employs a single frequency-selective LC branch and directly uses the two voltage outputs of the DOVCII to generate a notch response without additional active stages. Analytical expressions for the transfer function, notch frequency, and quality factor are derived, highlighting independent control of the passband gain and notch parameters. A sensitivity analysis demonstrates that the notch frequency depends exclusively on the LC product with half-order sensitivities, while the quality factor is predominantly controlled by a single resistor, resulting in predictable tuning and improved tolerance to passive component variations. Transistor-level analysis of the proposed filter was carried out using a standard AMS 0.35 μm CMOS process and has been validated through both circuit-level simulations and experimental measurements using a DOVCII implementation based on the AD844 current-feedback amplifier. Prototypes operating at 100 kHz and 50 Hz notch frequencies have been implemented, the latter employing a current-mode inductance simulator to avoid bulky passive inductors. Full article