Search Results (6,559)

Deep learning models are widely used in autonomous vehicle systems for perception, localization, and decision-making. However, their lack of transparency poses significant challenges in safety-critical environments. This systematic review presents a unified mathematical framework for explainable deep learning which integrates multimodal inputs, graph-theoretic road geometry, uncertainty modeling, and intrinsically interpretable representations. Road-structured priors that include lane topology and spatial constraints are incorporated into learning and optimization processes for ensuring model predictions and explanations to remain physically and semantically grounded. The review synthesizes methods across saliency-based, concept-based, causal, and intrinsic explainability, and extends them to vision-language models. This enables language-grounded, human-interpretable reasoning in autonomous vehicle systems. While vision-language models offer a new paradigm for semantic explainability, their limitations such as hallucinations, misgrounding, and reduced reliability under distribution shifts are also critically examined. Along with the role of road priors in improving alignment and robustness, another key contribution of this work is its quantitative evaluation metrics for road-aware explainability. These evaluation metrics link the explanations to spatial consistency, uncertainty alignment, and graph-structured reasoning. The overall framework connects latent representations, predictions, and explanations within a single formulation, enabling systematic comparison and analysis across models. Based on a PRISMA-guided review of 164 studies, this research identifies gaps in real-world reliability, temporal reasoning, and standardized evaluation, and outlines future directions including human-in-the-loop systems, regulatory readiness, and language-based auditing. Overall, this study advances a mathematically grounded and road-aware perspective on explainable vehicle AI which significantly bridges the gap between high-performance models and transparent, trustworthy autonomous systems. Full article

NanoArduSiPM represents a paradigm shift in the ArduSiPM (Architected Detection Unit for Silicon Photomultipliers) roadmap, evolving from a standalone instrument into a high-density modular building block (36 mm × 42 mm × 3 mm, 7 g). This revision does not merely pursue miniaturization; it re-engineers the signal-processing chain to maintain high performance within a scaled-down footprint, enabling the transition from single-unit detection to scalable, distributed multi-detector systems. NanoArduSiPM is based on a three-layer architecture comprising an external scintillator and Silicon Photomultiplier (SiPM) detection module, a dedicated high-speed discrete analog front-end, and a System-on-Chip (SoC) for embedded acquisition and processing. The physical implementation adopts high-integrity PCB routing and rigorous isolation techniques designed to suppress digital–analog coupling, a critical requirement in such a compact form factor. This deterministic layout strategy provides the architectural foundation for time-tagging capabilities, currently under quantitative characterization, by addressing the fundamental sources of signal interference at the hardware level. Beyond hardware integration, NanoArduSiPM introduces the capability for extended firmware functionality, including event tagging via external inputs and the implementation of coincidence and veto logic. This framework supports the acquisition of multiple correlated histograms and allows multiple units to be interconnected on a shared SPI bus. By shifting from standalone operation to a coordinated, hierarchical architecture, NanoArduSiPM enables distributed detection schemes where event selection and correlation are handled natively within the system, reducing the dependency on external data acquisition electronics. The compact modular architecture, together with the high-performance discrete analog front-end and embedded data handling, makes NanoArduSiPM suitable for applications where low mass and low power consumption are critical, targeting applications such as space-based payloads, laboratory instrumentation, remote sensing, and large-scale distributed multi-channel detection systems. While no radiation-tolerance qualification of the complete system has been performed in this work, the microcontroller family used in the design is also available in radiation-tolerant variants, which may support future implementations targeting more demanding radiation environments. Full article

GaN-on-Si light-emitting devices have been widely studied in the field of opto-electronics, while their optical performance and characterization accessibility are severely limited by the strong visible light absorption of the native silicon substrate. Conventional substrate transfer technologies often suffer from inherent thermal, optical, or mechanical bottlenecks. In this study, we developed a robust wafer-level substrate transfer strategy for 8-inch green GaN-on-Si light-emitting device wafers, utilizing a hybrid planarization process combined with SiO₂–SiO₂ direct bonding. The hybrid planarization precisely eliminated the 900 nm macroscopic steps, achieving sub-nanometer surface roughness for high-yield wafer bonding. We systematically investigated the physical evolution during substrate removal. Results indicate that the removal of the thick native silicon and high-stress buffer layers effectively released the additional in-plane biaxial compressive stress within the multiple quantum wells (MQWs), thereby mitigating the quantum-confined Stark effect (QCSE). Benefiting from the elimination of the light-absorbing silicon substrate and the incorporation of a built-in back-surface reflector (BSR), the transferred devices achieved a remarkable 1.9-fold enhancement in relative optical performance, albeit with an inherent trade-off of increased reverse leakage current while preserving basic diode functionality. Furthermore, optothermal dynamic analysis at high injection levels suggests a potential localized thermal bottleneck at the thick SiO₂ bonding interface, where a hypothesized heat-induced spectral red shift may counteract the carrier-screening blue shift. This work provides a feasible wafer-level substrate transfer process for GaN-on-Si devices and offers systematic experimental insights into stress relaxation and optothermal behaviors during the substrate transfer process. Full article

Recent advances in medical image segmentation have introduced a wide spectrum of deep learning paradigms, including 2D/3D convolutional neural networks (CNNs), transformer-based architectures, vision-language models (VLMs), prompt-driven foundation models such as Segment Anything Model (SAM), and diffusion-based approaches. Although these methods have demonstrated remarkable performance across MRI, CT, PET, ultrasound, and endoscopic imaging, the rapid proliferation of architectures has created methodological uncertainty regarding optimal model selection under varying clinical and data constraints. Existing surveys primarily focus on architectural categorization, yet provide limited guidance for decision-oriented model selection. This study presents a comprehensive and decision-guided survey that systematically analyzes segmentation paradigms across imaging modalities, task types, dataset characteristics, and evaluation protocols. Beyond taxonomy, we propose a practical model selection framework that links clinical scenarios, such as small lesion detection, multi-organ 3D segmentation, limited-data regimes, and domain shift, to appropriate segmentation strategies. Furthermore, robustness, generalization, annotation variability, and benchmarking reproducibility are critically examined. By integrating architectural taxonomy, cross-modal comparative analysis, and a structured decision framework, this work provides a clinically oriented roadmap for selecting segmentation methods and highlights future research directions toward reliable and reproducible medical AI systems. Full article

A wideband linearly polarized over-2-bit transmitarray antenna (TA) using the receiving-transmitting (R-T) scheme in the millimeter-wave band is presented in this work. The TA unit consists of two rectangular patches with a pair of bent branches, and the patches are connected by a metalized via. Two methods are used in this TA to obtain an over-2-bit phase shift of 0–${90}^{\circ }$ and 180–${270}^{\circ }$ from 18 GHz to 30 GHz. Firstly, ${180}^{\circ }$ phase resolution is obtained by rotating the receiving patch around via by ${180}^{\circ }$. Secondly, by tuning the connection position between the branches and rectangular patch of the TA unit cell, a continuous ${90}^{\circ }$ phase shift is further achieved. A TA prototype with $20\times 20$ units is designed, fabricated, and measured. The measured 1 dB and 3 dB gain bandwidth is 24.9% (24.47–31.43 GHz) and 46.96% (20.45–33 GHz) respectively, with a peak gain of 25.17 dBi and a peak aperture efficiency of 55.2%. The measured results agree well with the simulated ones. Full article

Real-world time series forecasting is often challenging due to non-stationarity and distribution shifts, where the optimal forecasting model varies across different temporal regimes and horizons. In this work, we introduce a method called Adaptive Future-Guided Ensemble Learning (AFG-EL), a two-stage framework that performs drift-aware, sample-level routing over a heterogeneous model zoo. AFG-EL learns dynamic fusion weights from meta-features of the historical window and incorporates a future-guided training signal from a relative-future teacher or scorer, emphasizing learning on regime transitions and drift-sensitive segments. Crucially, the inference process remains strictly causal, requiring only historical data and extracted meta-features. We further use sparse routing with an entropy-based fallback mechanism to enhance stability when routing confidence is low. Our experiments on several commonly used forecasting datasets demonstrate that AFG-EL consistently outperforms strong single-model baselines, uniform averaging, and adaptive fusion baselines. Full article

Proton exchange membrane water electrolysis is a promising technology for large-scale green hydrogen production due to its high efficiency, compact design, and rapid dynamic response. However, its commercialization is strictly limited by high material costs, durability issues, and complex multiphysics coupling within the membrane electrode assembly. This work provides a comprehensive and critical review of key physicochemical processes and advanced predictive modeling approaches for PEMWEs. To capture recent paradigm shifts, we introduce an innovative multi-dimensional classification framework—incorporating spatial resolution, temporal dynamics, and methodological paradigms—to critically evaluate lumped-parameter, continuum, microscale, and multiscale models, explicitly defining their applicability bounds and inherent limitations. The fundamental mechanisms governing electrode kinetics, membrane water transport, and gas–liquid two-phase flow are analyzed, establishing state-of-the-art quantitative benchmarks for microstructural parameters and advanced 3D flow field topologies under high-current-density and high-pressure regimes. Furthermore, we systematically examine model validation rigor, typical prediction errors, and the critical failure of static models in capturing dynamic property shifts during extreme bubble breakthrough. Recent breakthroughs integrating in situ diagnostics, pore-scale simulations, density functional theory, and Physics-Informed Neural Networks are extensively discussed. Future efforts must prioritize mechanical–electrochemical–thermal coupling, transient degradation prognostics, and machine learning-driven predictive digital twin technologies to overcome current empirical limitations and accelerate the gigawatt-scale deployment of PEMWE systems. Full article

Sea buckthorn (Hippophae rhamnoides L.) leaves are rich in nutrients and bioactive constituents, with great potential for fermented tea development. It has been demonstrated that Fuzhuan brick tea processing can improve sea buckthorn leaf tea flavor, but the underlying microbial succession remains unexplored. Therefore, we characterized the dynamic succession and interrelationships of bacterial and fungal communities via Illumina NovaSeq 6000 sequencing. β-diversity analysis revealed successive shifts in microbial community structure, with fungal communities changing mainly in the early stage and bacterial communities varying more in the late stage of fermentation. The relative abundance of Pseudomonas, a genus frequently associated with flavor formation and tea quality, increased steadily. Fungal taxonomic analysis revealed that the genus Aspergillus, particularly the species Aspergillus chevalieri, remained dominant throughout the fermentation process. Linear discriminant analysis effect size indicated an enrichment of microbial taxa typical of fermentation, accompanied by a relative reduction in putative opportunistic microbes. Additionally, Aspergillus exhibited significant negative correlations with five key differentially abundant bacterial genera. Interestingly, microbial co-occurrence networks suggested an overall tendency toward coexistence rather than mutual exclusion between the bacterial and fungal communities. This work provides a theoretical foundation for the development of novel fermented sea buckthorn leaf tea products. Full article

Due to the curse of dimensionality faced in modern industrial processes, high-dimensional Statistical Process Control (SPC) faces significant challenges in detecting small and sparse process shifts. Traditional multivariate control charts often suffer from noise accumulation and fail at timely identification of anomalies that affect only a small subset of variables. To address this issue, this study proposes an enhanced Multivariate Exponentially Weighted Moving Average (MEWMA) approach with variable selection and adaptive sampling for efficient process monitoring. The proposed smart approach works in two ways: first, it automatically focuses on the variables that are most likely to have changed (variable selection); second, it takes samples more frequently when things look uncertain, and less frequently when everything appears stable (variable sampling interval). This combination allows problems to be detected earlier. A Monte Carlo approach is used to calculate the the Average Time to Signal (ATS) values of the proposed scheme, and comparative results show that the proposed scheme outperforms standard charts like the Fixed Sampling Intervals (FSI) VSME, VSI-$T^2$, and VSI-MEWMA schemes in terms of detection speed for small-to-moderate sparse shifts. Finally, a real example from car body manufacturing is provided as an illustration for the implementation of the proposed scheme. Full article

Proof-of-Work (PoW) cryptocurrency mining is conventionally characterised as an energy competition, yet this paper provides evidence that the primary competitive margin has shifted from electricity procurement to semiconductor acquisition. Using Bitcoin (BTC) and Bitcoin Cash (BCH)—two SHA-256 networks sharing identical hardware but differing in scale and governance—as a natural comparative setting, we apply the Autoregressive Distributed Lag (ARDL) bounds testing approach to 112 weekly observations (January 2019–March 2021). Mining reward exhibits near-unity long-run elasticity with respect to both hash rate and energy consumption (0.773–1.009), confirming miners’ price-taking behaviour. Critically, the shutdown threshold—an efficiency-based cost floor derived from ASIC hardware generations—dominates all cost-side regressors with elasticities of $-1.941$ to $-2.156$, substantially exceeding electricity price effects in both magnitude and significance. VAR analysis provides evidence consistent with a centralisation paradox: rising chip efficiency Granger-predicts increased mining pool concentration for BTC (${\chi }^2=33.64$, $p<0.001$) via a revenue-redistribution mechanism, while electricity costs carry no equivalent structural consequence. Zivot–Andrews tests confirm that China’s 2021 mining ban produced a significant transient disruption but no permanent structural break in BTC’s hash rate trajectory, consistent with the geographic mobility of capital-intensive hardware. These findings imply that standard energy-price policies address the wrong margin; effective governance of PoW sustainability requires redirecting regulatory attention toward the semiconductor supply chain—a conclusion with direct relevance to SDG 7 and SDG 13. Full article

Generative artificial intelligence (AI), especially large language models (LLMs) that can write and debug code, is changing how students approach programming work in engineering education. Unlike more open-ended conceptual or modeling tasks, programming fits closely with what these systems do well: generating syntax, fixing errors, building procedural logic, and completing code structures. Hence, programming coursework may be one of the areas in which AI changes performance patterns in a measurable way. This study examines whether that shift appears in actual student outcomes. Using a retrospective pre/post design, it compares results from a pre-AI period (2021–2022) with results from a post-AI period (2023–2025), when generative AI tools became widely available to students. The focal assessment is a comprehensive programming project graded with the same rubric across multiple sections and terms. Performance is evaluated through descriptive statistics, distributional comparisons, and mastery thresholds (≥80%). The post-AI period shows a rise in overall scores, along with strong clustering near the top of the scale. Lower- and middle-range scores become much less common, most students fall in the highest score band, and overall variability declines. These results suggest that generative AI acts as a procedural equalizer in programming contexts, referring to the role of generative AI in reducing performance differences by assisting with rule-based, syntax-driven, and execution-oriented aspects of tasks, thereby raising baseline outcomes while compressing variation among students. It appears to raise lower-end performance and make outcomes more consistent, but it also narrows the spread among stronger students and creates a ceiling effect. That pattern raises questions about assessment validity, skill differentiation, and what “mastery” means when AI can handle much of the procedural work. Using multi-term data from authentic online courses, this study adds empirical evidence to the growing literature on AI in engineering education and identifies programming coursework as a setting where generative AI may have already changed performance dynamics in a structural way. Full article

The Lhalu Wetland National Nature Reserve, the largest natural urban wetland on the Qinghai–Tibet Plateau, plays a critical role in maintaining regional ecological balance and biodiversity. However, the baseline biodiversity of this reserve remains unclear because of the extensive temporal span of historical records, shifts in taxonomic systems, and inconsistent survey methodologies, which impedes a robust scientific understanding of its ecological dynamics. This study systematically compiled and taxonomically verified species records from over 50 sources spanning the 1950s to the present. The records cover plants, fish, birds, and amphibians/reptiles, thereby resolving issues of synonyms, homonyms, and misidentifications. Each species record is annotated with its original survey time, allowing users to distinguish historically reported occurrences from those recorded in recent surveys. Species accumulation curves were constructed for major taxa and compared with 45-year climatic trends (1979–2023) and socioeconomic indicators for Lhasa City. A total of 438 vascular plant species (82 families, 251 genera) and 311 animal species (39 orders, 98 families), including 30 fishes, 174 birds, and 11 amphibians/reptiles, were documented. Invasive species comprised 55 alien plants and 13 alien fishes, while 4 plant and 46 animal species are under national protection. Temporal synchrony between increases in alien taxa and anthropogenic pressures (gross domestic product (GDP) and population growth, infrastructure development) suggests that human activities may be a potential driver of biodiversity change, but formal causal inference is precluded by heterogeneity in survey methods and sampling effort. This work provides a structured dataset of the biodiversity baseline of the Lhalu Wetland and offers a descriptive assessment of its temporal patterns in relation to climate and human disturbance, while explicitly acknowledging data limitations. It provides essential data and theoretical support for the scientific management and targeted conservation of plateau urban wetlands. Full article

Taking J.R.R. Tolkien’s portrayal of mercy in The Lord of the Rings as a point of departure, this article examines a question long debated in Dharmic commentarial traditions: what are the roles of individual effort and grace in completing the path to God-realization? The Indian spiritual teacher Meher Baba (1894–1969) offers a cosmology in which consciousness evolves by winding impressions (saṃskāras) through millions of lifetimes and progresses by unwinding them in thousands more, yet cannot complete this unwinding through effort alone. The final wiping out of all impressions requires the grace of a Sadguru or God-realized “Perfect Master.” This necessity is structural on two grounds, both rooted in the nature of consciousness itself. Building on Murshida Carol Weyland Conner’s distinction between “ascendant” and “descendant” paths of God-realization, this article examines what Meher Baba claimed to accomplish as Avatar: the cutting of a hole through which unprecedented divine light descends into physical creation. The descendant epoch inaugurated by this work shifts the orientation of incarnate existence from liberation out of matter toward progressive perfection within it. The wheel of rebirth is not abolished in this view. Through the Avatar’s wounded body, it is wounded into a new form, its substrate becoming divinized matter and its telos becoming perfection. Grace operates not only at the threshold of individual liberation but throughout the field of reincarnation itself. Full article

This work presents a noninvasive imaging method to locate veins using a tuned microwave loop resonator. It offers a low-cost, fast, and effective solution to the challenges in venipuncture. The sensor features a loop resonator with a 5.2 mm radius, incorporating a self-tuning mechanism, and operates at 2.408 GHz with a reflection coefficient of −48.77 dB. It generates localized high-intensity electric fields that penetrate tissues to sufficient depths, enabling the detection of veins based on shifts in resonant frequencies that are induced by the varied dielectric properties of blood vessels. Two-dimensional raster scan simulations of the cephalic and median cubital veins yielded a ∼25 MHz downward resonant-frequency shift between vein and non-vein positions, with the median cubital vein still detectable at depths up to 6 mm. To quantify generalization to real tissues, a decision tree classifier trained on 63 simulation samples and evaluated on 335 in vivo measurements achieved 82.09% classification accuracy (sensitivity 81.25%, specificity 83.02%), demonstrating that the simulation-derived frequency contrast transfers reliably to experimental data despite inter-subject tissue variability. Extensive tests conducted demonstrate the sensor’s effectiveness, producing consistent and distinguishable frequency shifts when the sensor moves on the skin across veins. This technology holds significant promise for improving venipuncture accuracy, minimizing complications, and enhancing patient comfort. Full article

In magnetron sputtering-based gas aggregation sources, nanoparticle formation and yield are strongly influenced by the flow-regulated transport and residence time of particles within the condensation chamber. However, the effect of internal geometric parameters on flow structure and nanoparticle growth is not well understood. In this study, computational fluid dynamics (CFD) coupled with a discrete phase model (DPM) is employed to investigate how magnetron radius affects flow characteristics, particle transport, and their implications for nanoparticle formation. The results show that increasing the magnetron radius significantly enhances axial flow uniformity and suppresses vortex structures near the inlet. This shift from radial diffusion-dominated to primarily axial transport effectively reduces particle trapping and wall deposition. Furthermore, the regulation of flow structure modifies particle residence time distributions, which is considered a key factor associated with nanoparticle growth potential and size evolution in gas-phase synthesis. Larger magnetron radii promote more stable transport pathways and improve particle transmission efficiency, thereby improving particle transmission efficiency and providing more favorable conditions for nanoparticle formation. These findings indicate that geometric optimization can simultaneously enhance transport efficiency and influence the conditions potentially favorable for particle growth, providing valuable guidance for the design of high-yield nanoparticle synthesis systems. Overall, this work provides insight into how flow field characteristics influence nanoparticle transport and potential growth behavior, offering a foundation for optimizing magnetron sputtering-based nanoparticle synthesis. Full article