With the deep integration of renewable energy and DC loads, DC microgrids have become critical components of modern power systems. The widespread application of software-intensive controllers and power electronic devices has enhanced system flexibility but has also notably increased the risks of cyber-attacks and actuator faults. Such risks may compromise system stability and, in severe cases, cause damage to critical equipment, resulting in substantial economic losses. To address these issues, this paper proposes a distributed adaptive fault-tolerant control strategy based on fixed-time consensus and deep reinforcement learning (DRL), aiming to enhance the operational reliability of DC microgrids under actuator fault conditions. Firstly, the actuator fault effects are mapped to corruptions of control signals, which are adaptively compensated by voltage and current secondary control. Then, current sharing and voltage recovery mechanism under actuator faults is proposed based on fixed-time consensus and DRL algorithms. Besides, the fixed-time stability of the proposed control strategy is rigorously proven through the Lyapunov function. Simulation and control-hardware-in-the-loop (CHIL) experiments demonstrate that the proposed strategy exhibits excellent fault-tolerant capability and robustness under various actuator faults, load changes and communication interruptions.

Recent electrocardiogram (ECG) foundation models have extended ECG analysis from rhythm and morphology recognition to cardiac structure-function assessment, but the computational and memory requirements limit their use in resource-constrained settings. This study aimed to reduce the size and inference cost of a unified ECG foundation model while retaining its multitask performance and clinically relevant representations. We developed the ECGFM-U-Lite, a compact 1D-MobileViT student, using a three-stage compression procedure. Task outputs and intermediate representations were firstly transfered from teacher through structure-function consistent distillation, redundant channels were than removed by using physiologic attention-guided pruning while retaining anatomically informative pathways, and the structure-function-aware quantization strategy achieves adaptive precise quantification by leveraging the sensitivity discrepancies across different modules finally. The model was evaluated on six locked external cohorts. Its computational profile was examined on a laptop GPU and under Android- and Jetson-Nano-oriented ONNX Runtime simulations. ECGFM-U-Lite achieved an overall AUC of 0.862, compared with 0.918 for the teacher, and retained 93.8% and 93.5% of teacher AUC for rhythm/morphology and structure/function tasks, respectively. ECG-to-Text and Text-to-ECG R@1 values were 0.766 and 0.762 in the teacher-derived semantic consistency probe. The parameter count decreased from 86.4 M to 655.7 K, model size from 345.6 MB to 1.7 MB, and peak RAM from 320 MB to 4.4 MB. The inference process was 3.6-fold faster on the laptop GPU and approximately 3.7-fold faster in the simulated edge-oriented profiles. U-Lite substantially reduces the resource requirements of a unified ECG foundation model while preserving most of its external discrimination and teacher-derived cross-modal representation structure, providing a practical basis for subsequent evaluation on resource-constrained hardware.

Soybean protein content and fungal contamination level are critical determinants of economic value and consumption safety. However, conventional analytical methods rely on destructive chemical analyses or single-parameter measurements, severely limiting assessment efficiency and multi-attribute evaluation. Although near-infrared spectroscopy (NIRS) is widely used, it frequently suffers from spectral redundancy, model instability, and limited generalization in complex agricultural matrices. To address these challenges, this study proposes a parallel NIR analytical framework driven by Ensemble Monte-Carlo Stability-based Uninformative Variable Elimination (EMS-UVE). EMS-UVE integrates repeated stability estimation with a noise-referenced adaptive threshold for high-efficiency informative wavelength selection. For fungal contamination classification, EMS-UVE-XGBoost reduced the feature subset to 12 wavelengths while improving the validation accuracy and macro-F1 score to 0.9529 and 0.9530, respectively, outperforming the baseline ensemble Monte Carlo uninformative variable elimination-based extreme gradient boosting (EMCUVE-XGBoost) (21 wavelengths; accuracy = 0.9412, macro-F1 = 0.9418). For protein content regression, EMCUVE-SVR retained 18 wavelengths, with Rcv=0.8834, RMSECV = 1.6027%, Rp=0.8875, and RMSEP = 1.3338%, whereas EMS-UVE-SVR retained 13 wavelengths, achieving superior performance (Rcv=0.9012, RMSECV = 1.4720%, Rp=0.9070, and RMSEP = 1.2042%). Across 50 independent runs, EMS-UVE-XGBoost yielded a mean validation accuracy of 0.9518 ± 0.0062 and a macro-F1 score of 0.9516 ± 0.0068, while EMS-UVE-SVR achieved af (Rp) of 0.9055 ± 0.0108 with an RMSEP of 1.219 ± 0.058%, demonstrating exceptionally low run-to-run variability. Furthermore, the finalized dual-task models were successfully deployed on a Raspberry Pi 5-based embedded platform. External validation confirmed that the system completed shared spectral acquisition and simultaneous multi-attribute inference within an end-to-end measurement cycle of approximately 5.6 s, achieving a classification accuracy of 0.9200 and a protein prediction correlation coefficient of 0.9067. These findings demonstrate an effective, robust, and edge-deployable strategy for rapid, non-destructive, and multi-attribute evaluation of complex agricultural products.

Tibial stress fractures are overuse injuries in runners and are associated with cumulative internal bone stress. Current wearable devices usually measure external biomechanical loads rather than tissue-level stress within the tibia. We developed and evaluated a wearable-sensor-based surrogate model for reconstructing the temporal evolution of tibial bone stress in healthy male recreational runners, using participant-specific dynamic finite element analysis (FEA) as the supervision source. Dynamic FEA-derived stress fields from 50 male runners, with heterogeneous and orthotropic material properties, were used as ground-truth regression targets for deep learning. The main model used synchronized inertial measurement unit and surface electromyography signals as inputs and a convolutional-transformer backbone to map surface signals to dynamic tibial stress outputs. A self-correction with adaptive mask training module was also evaluated as an auxiliary regularization strategy. The model predicted temporal stress waveforms (R2 = 0.90) and global peak magnitudes (Pearson r = 0.944), with low systematic bias but residual individual-level variability. Robustness training was evaluated under sensor noise and modality-loss conditions, and edge-device inference required less than 60 ms. These findings indicate that wearable sensors can support tibial stress reconstruction in male recreational runners, although external validation in female runners and other populations is required before broader application.

Wound healing and tissue reconstruction following a skin defect, particularly the restoration of full skin function and appearance, remain a major clinical challenge. Traditional hydrogel dressings face limitations in mechanical strength, poor permeability, and a lack of multifunctionality, making them inadequate for meeting the dynamic demands of the wound healing process. Fibrous hydrogels offer a compelling alternative by combining the high-water content and biocompatibility of hydrogels with the anisotropic structure and design flexibility of fibers, allowing them to closely mimic native tissue architecture. In this review, we introduce a three-tier framework based on structural, functional, and responsive programmability to systematically classify and evaluate programmable hydrogel fibers for wound healing. Recent advances in structural design strategies, functionalization approaches, and biomedical applications are summarized. Notably, this review uniquely integrates structural design, functional strategies, and translational potential within a unified programmability framework, while critically examining current limitations and challenges in clinical translation. Despite these obstacles, hydrogel fibers hold considerable promise for personalized tissue engineering and functional regeneration.

Unlike specialized breeds, High-Andean Creole cattle have developed a remarkable adaptability and hardiness to challenging high-altitude conditions, albeit at the expense of productivity. However, the efficiency of their reproductive performance under estrus synchronization protocols for fixed-time artificial insemination (FTAI) remains scarcely addressed. This study aimed to evaluate the response of Creole cows to two FTAI protocols in the districts of Chuschi (Cangallo) and Chipao (Lucanas), Ayacucho, Peru. From a sample of 60 Creole cows (identified by phenotype and coat color), with uniform body condition scores and management system, 35 were synchronized using a three-handling protocol (T1: estradiol cypionate and Bovine Intravaginal device (BID) removal on day 8), and 25 using a four-handling protocol (T2: BID removal on day 7 and estradiol benzoate on day 8). The cows were inseminated using thawed semen from Fleckvieh and Brown Swiss bulls, the latter belonging to the genetic nucleus of the Instituto Nacional de Innovación Agraria. Estrus detection rates ranged from 88.6% to 80.0%, yielding pregnancy rates of 57.1% and 56.0% for T1 and T2, respectively (p>0.05). No significant effects of protocol, age, number of calvings, body condition score, or bull breed were found on estrus detection, pregnancy, or calf sex (p>0.05). However, cows with a corpus luteum had 7.1-fold higher odds of pregnancy than those presenting only follicles (p<0.05). In conclusion, both FTAI protocols demonstrated comparable reproductive performance; however, the three-handling protocol with estradiol cypionate emerges as the preferred option due to its lower cost and greater operational simplicity under challenging High-Andean conditions.

Lysosomes are degradative, metabolic and signaling organelles in central nervous system (CNS) cells and play essential roles in neural homeostasis, injury responses and disease progression. Changes in lysosomal function are required for neurons, microglia, astrocytes and oligodendrocyte-lineage cells to adapt to proteotoxic stress, lipid overload, mitochondrial damage and inflammatory stimulation. Therefore, lysosomal activity in the CNS is regulated by diverse chemical and cellular cues, including luminal acidification, hydrolase maturation, ion transport, membrane integrity, lipid handling and transcriptional lysosome biogenesis. Here, we summarize current knowledge of lysosomal organization and function in neural cells, including lysosomal membrane signaling, vacuolar H+-ATPase (V-ATPase)-dependent acidification, ion and redox regulation, lysosomal membrane quality control, and metabolic recycling of proteins, lipids, glycans and damaged organelles. We then survey pathological lysosomal remodeling in spinal cord injury, traumatic brain injury, Alzheimer's disease, Parkinson's disease and related neurological disorders. Across these conditions, lysosomal remodeling may initially support cargo clearance and stress adaptation, but becomes maladaptive when degradative capacity, membrane repair or lipid-processing ability fails to meet cellular demand. Finally, we discuss lysosome-oriented therapeutic strategies for neural repair, including restoration of lysosomal acidification and hydrolase activity, transcription factor EB/transcription factor E3 (TFEB/TFE3)-mediated lysosomal biogenesis, regulation of lysosomal ion channels, enhancement of myelin lipid turnover, and lysosome-responsive nanomedicine. This Review highlights lysosomes as active organizers of CNS injury and repair rather than passive endpoints of degradation.

Metaholograms have attracted significant attention for their ability to project high-quality images within a compact form factor. Although various multiplexing strategies have been explored for multi-channel metaholograms, current approaches still face limitations in display channel extension and inter-channel crosstalk elimination. Here, we introduce a waveguide-based metahologram that enables four independent, crosstalk-free display channels via diffraction-order multiplexing. The metaholograms are integrated onto the top surface of a flat glass waveguide and modulate guided light through detour phase. With carefully designed device parameters, light diffracted into four distinct orders is selectively out-coupled to form crosstalk-free holographic images. To decouple the detour phase modulations between different orders, we develop an adaptive-weight phase-retrieval algorithm that simultaneously optimizes phase profiles across multiple diffraction orders to realize independent image encoding. Extending this strategy, we also demonstrate a six-channel metahologram for two full-color image projection, highlighting the versatility and scalability of the design. This work offers a new pathway toward compact, high-capacity holographic devices, with promising applications in AR/VR displays, information storage, and image encryption.

Tactile sensing in biological systems is a sophisticated spatiotemporal process where mechanoreceptors decode surface properties through rhythmic interactions. However, current artificial tactile recognition predominantly relies on discrete classification under highly constrained conditions, making it difficult to capture the quantitative physical dimensions of stimuli in naturalistic settings. This study proposes a spatiotemporal method for the quantitative identification of regular texture spacing under naturally varying tactile interaction conditions, bridging the gap between raw sensor dynamics and geometric perception. We utilized a prosthetic finger integrated with a piezoresistive sensor array to capture tactile responses during dynamic sliding. To characterize the frequency-tuning of the sensor array and the energy distribution of rhythmic excitations, Spectral Centroid Analysis (SCA) was employed for initial feature extraction, revealing a spatial coupling effect during tactile interaction. Subsequently, a hierarchical deep learning model was developed, combining Convolutional Neural Networks (CNN) to extract localized spatial contact patterns and Long Short-Term Memory (LSTM) units to integrate quasi-periodic temporal modulations. A channel-wise attention mechanism was incorporated to adaptively weight the contributions of non-uniform sensing regions. Experimental results demonstrate that the proposed approach achieves a recognition accuracy of 76.80% in identifying texture spacing, while indicating improved stability compared to baseline models that lack spatiotemporal integration. Crucially, interpretability analysis reveals that the attention mechanism effectively prioritizes informative sensing channels amidst the signal variability induced by unconstrained force. Our work establishes a attention-based approach to tactile decoding, demonstrating that the coupling of spatial heterogeneity and temporal periodicity is essential for robust perception. This approach provides a computational foundation for developing advanced neuro-prosthetics and autonomous robots capable of human-like tactile discernment in unconstrained environments.

High-brightness-integrated quantum light sources represent a fundamental element for large-scale photonic quantum technologies. Integrated microring resonator (MRR) in highly nonlinear and ultralow-loss material platforms enable entangled photon-pair generation, but device performance is affected by fabrication tolerances and can thus not be judiciously controlled. Redundant designs based on parameter variation are impractical for chip-scale circuits, as the sheer number of parameters would be too complex. To address this, passive photonic circuits must be combined with active control elements to optimize photon-pair emission and to match emission characteristics among sources. Here, we present a comprehensive study of an actively tunable, interferometrically coupled integrated MRR implemented in silicon nitride (SiN). By manipulating two thermally controlled phase shifters, the ring-bus coupling and the resonator phase can be dynamically adjusted, enabling post-fabrication control of its quality factor (Q-factor) between 105 and 106 to optimize the photon emission bandwidth and the pair generation rate. We directly demonstrate for the first time that the optimal coupling and emission characteristics do not occur at critical coupling and vary across frequency modes due to dispersion effects, in good agreement with our theoretical model. These findings highlight the role of adaptive integrated photon sources for single-batch fabrication of reconfigurable devices with application-specific spectral properties.

Multispectral stealth is a critical technology in modern military and security applications. Conventional approaches generally address stealth in only one or two spectral bands-such as radar, visible light, or infrared (IR)-thereby limiting their effectiveness against advanced detection systems. In this study, we present, for the first time, an integrated and dynamically adaptive electrochromic system capable of achieving simultaneous stealth across the radar, near-infrared (NIR), and visible spectra. This single-device platform surpasses traditional independent or discrete methods in complexity, efficiency, and real-time tunability. Our multispectral stealth device employs a vertically integrated electrochromic device (ECD) that unifies stealth functionality across all three spectral domains. Visible and NIR camouflage are realized through voltage-controlled, real-time modulation of reflectance, enabled by the electrochromic properties of the ECD and its embedded indium tin oxide (ITO) layers. Radar camouflage is achieved through precisely patterned ITO structures incorporated within the device, effectively suppressing electromagnetic wave reflections. To validate this approach, we performed numerical simulations and experimental characterizations using an X-band waveguide, confirming consistent and robust multispectral stealth performance. This work represents a significant advancement in multispectral stealth technology by integrating radar, NIR, and visible light camouflage into a single, tunable platform.

With the increasing demand for tunable optical functionalities, metasurfaces capable of dynamic wave control have attracted growing attention. Compared with reconfigurable designs that rely on external stimuli, incident-light-modulated multiplexed metasurfaces offer a unique scalable pathway for achieving tunable functionalities. In this work, based on multiplexed metasurface schemes, we show hybrid multiplexing of polarization, wavelength, and orbital angular momentum (OAM) can be applied for dynamic tuning. We start to demonstrate a noise-engineered metasurface platform that supports dynamic holographic display across multi-nonorthogonal polarization channels, further combined with position, wavelength, and OAM multiplexing. Moreover, we extend this approach to achieve simultaneous multiplexing of polarization, wavelength, and OAM. This integration enables a single device to dynamically switch among holographic patterns under different incident light configurations, significantly enhancing the functional diversity of dynamic displays. Our strategy presents a highly compact and efficient solution for high-capacity multi-dimensional holography, with promising applications in advanced displays, data storage, optical encryption, and adaptive photonic systems.