All-solid-state batteries (ASSBs) are viewed as a consequential step in the development of rechargeable batteries for enabling higher energy and power densities, particularly with sulfide-based solid electrolytes (SEs). Contrasting these benefits, their low resistance to mechanical loads introduces productionrelated challenges during cell assembly,especially component handling. The present research investigates the handling of sulfide-based SEs (Lithium phosphorus sulfur chloride) in ASSB cell stacking with various state-of-the-art grippers to identify optimization potentials. The experimental design systematically evaluates key parameters such as holding force, gripper speed, and deposition distance, in relation to the achievable deposition accuracy per ISO 9283. All tested grippers achieve deposition accuracies of less than 0.2 mm or 0.2° in 74% of parameter settings. The highest accuracies are observed with a vacuum suction gripper with a Polytetrafluoroethylene contact surface,operated at defined process conditions. These insights benefit the further development towards the accurate and secure handling within future ASSB manufacturing.

The escalating prevalence of ransomware attacks necessitates the development of innovative detection methodologies. Traditional detection techniques, such as signature-based and behavioral analysis methods, often fall short in identifying novel or obfuscated ransomware variants. To address these challenges, the concept of Semantic Flow Mapping has been introduced, offering a novel approach to ransomware detection. This methodology employs graph theory to model system processes and their interactions, representing processes as nodes and communication channels or resource accesses as edges. By applying statistical measures, the system assesses normalcy within these graphs, identifying anomalies through deviations from established baselines. Machine learning techniques, including clustering algorithms and neural networks, are utilized to model the complex, non-linear relationships characteristic of system behaviors, thereby enhancing the detection of ransomware activities. The mathematical framework is designed to be computationally efficient, ensuring real-time applicability without imposing significant performance overheads on the system. Empirical evaluations demonstrate the method's proficiency in maintaining low false positive and false negative rates, showing its reliability in distinguishing between legitimate and malicious processes. Furthermore, scalability assessments reveal that the methodology can be effectively deployed across networks of varying sizes, maintaining consistent performance without imposing substantial computational overhead.

Human movement emerges from the interplay between nervous, muscular, and skeletal systems, interacting with the environment. Understanding these processes is crucial for developing wearable robotic technologies to restore movement following neuro-muscular injuries. Movement neuro-mechanics is often studied via computer models of the composite neuromusculoskeletal system, which use static, dynamic optimization or reinforcement learning to estimate muscle activation and resulting mechanical forces from kinematic and kinetic data. However, such approaches often fail at capturing the variability in multi-muscle neural recruitment and force generation across movements, anatomies, and conditions (i.e., ageing or injury). Electromyography (EMG)-driven musculoskeletal modeling uses measured EMGs and joint angles for simulating muscle-tendon level mechanics with no assumptions on how muscles are recruited by the central nervous system. EMG-driven models enabled task-agnostic, myoelectric model-based controllers for bionic limbs and exoskeletons. However, real-time neuromechanical models still remain largely proprietary, hindering their widespread use, progress and standardization. Here, we introduce CEINMS-RT, a freely available, open-source, neuromechanical modeling framework for the real-time myoelectric model-based control of wearable robots, including exoskeletons, exosuits, haptic devices, and bionic limbs. CEINMS-RT explicitly models person-specific movement neuro-mechanics and estimates EMG-dependent variables including muscle activation, muscletendon force, and resulting joint dynamics. This represents an open-source alternative to end-to-end neural regressors, which do not estimate intermediate biomechanicasl variables that would be critical for roboust wearable robot control (e.g., joint stiffness, damping or underlying muscle-tendon impedance). Here, we introduce the CEINMS-RT framework and provide application results in the context of wearable robotic control.

While innovation can lead to groundbreaking products, it also typically entails higher costs due to the need for advanced design, production processes, and licensing fees for critical technologies. Therefore, RISC-V is crucial for cost-effective applications as a practical and robust instruction set architecture (ISA). RISC-V is a free-available and open-source instruction set architecture, allowing you to create CPU architectures that are not subject to any intellectual property rights. Due to the growing demand for IoT devices, it is more practical to create boards RISC-V specifically designed for IoT applications. Although significant progress has been made in creating open source CPU cores, more emphasis must still be placed on special abilities for IoT applications. Most open-source cores have inadequate documentation or are designed for specific purposes. This analysis focuses on multiple open-source CPUs and System-on-Chip (SoC) designs that leverage the RISC-V architecture, examining their relevance and effectiveness in IoT environments. The study further outlines a RISC-V SoC specifically designed for IoT applications and investigates the primary features of a suitable open-source IoT node, also identifying the need for ongoing R&D to support the development of specialized protocols.

This investigation yielded three conclusions. To begin, various counterexamples were constructed to establish the main theorem, which asserts that the optimal scaling factor for a model implemented with a self-attention function to converge and achieve its best possible accuracy is not 1 √ d k. Secondly, a theoretical analysis showed the existence of the kinetic energy that was obtained and stored in the model weights. Several empirical facts were discovered that support this theoretical interpretation. As a result of this discovery, a self-attention model, such as transformers, would have an inertial effect to memorize its original value of the scaling factor. As a result, once the optimal scaling factor is discovered early in training, it may be fixed for the subsequent training phase. Finally, the first result and second findings imply that this scaling factor is independent of the variance or mean of each value of entries of the input, query, and key matrices (based on the an observation of counterexamples and a proof provided to show self-attention function is a stack of Ising models), and the scaling factor represented a form of kinetic energy (based on the inertial effect). The first two results also confirm the conjecture that the best scaling factor is determined by the entire topology of the neural net function.

Handling hierarchical relationships in linguistic data has long been a challenging task, requiring advanced frameworks capable of dynamically processing complex contextual dependencies. By employing weighted aggregation and recursive propagation strategies, the architecture balances local context with global coherence, demonstrating adaptability across a wide range of linguistic tasks. Experiments revealed that structured datasets benefited significantly from the cascading design, while datasets with fragmented or informal text posed challenges that suggest potential areas for refinement. Resource usage analyses indicated proportional increases with cascading depth, with intermediate layers playing a critical role in enhancing performance. Error patterns across tasks such as parsing and classification highlighted opportunities for targeted improvements in specific linguistic domains. Evaluation of input length variations revealed the mechanism's ability to handle shorter sequences effectively while identifying sensitivity to extended contexts. The results showcased the architecture's capacity to generalize semantic relationships and adapt to diverse linguistic structures. Layer-wise ablation studies emphasized the contributions of intermediate refinements, showing the need for balanced depth in hierarchical modeling. The cascading mechanism's innovative framework sets a benchmark for future explorations, aligning computational demands with enhanced representational accuracy. Combining efficiency with versatility, the Neural Contextual Cascade Mechanism redefines approaches to hierarchical contextual modeling, offering a pathway for scalable linguistic applications.

The Predictive Code Anomaly Signatures (PCAS) framework represents a significant advancement in the proactive detection of ransomware threats. By integrating static code analysis with sophisticated anomaly detection algorithms, PCAS effectively identifies malicious code patterns indicative of ransomware activity. This innovative approach models code features as multi-dimensional vectors, capturing both syntactic and semantic characteristics to distinguish between benign and malicious code segments. The framework's scalability and modular adaptability facilitate seamless integration into diverse computing environments, enhancing its practical applicability. Empirical evaluations demonstrate that PCAS achieves a high true positive rate and a low false positive rate, indicating its robustness in accurately detecting ransomware threats. Furthermore, the ability of PCAS to bridge static and dynamic analysis paradigms addresses limitations inherent in traditional detection methods, offering a comprehensive solution for preemptive ransomware identification. The implementation of PCAS in cybersecurity infrastructures has the potential to significantly enhance defenses against evolving ransomware threats, providing real-time detection capabilities and enabling prompt responses to mitigate the impact of such incidents. Overall, the PCAS framework represents a substantial contribution to the field of cybersecurity, offering a novel and effective approach to ransomware detection.

This paper introduces a theoretical design for a secure server system aimed at mitigating hacking risks, particularly for sensitive applications like banking systems. The proposed design utilizes a layered architecture comprising sub-servers, command processors, and offline update mechanisms to ensure data integrity and security. A key feature is the integration of an AI Detection Restoring Device that continuously monitors server patterns, detects unauthorized changes, and identifies unusual or repetitive commands. The AI-driven system enhances security by reducing human error, automatically resetting interrupted flows, and ensuring data protection. Only the public server is connected to the internet, with other components remaining offline to minimize vulnerabilities. This approach prevents unauthorized tampering and ensures robust data protection.. The proposed system's applications extend to banking and privacy-focused platforms, providing high levels of security and user authentication. By integrating AI with multiple verification layers and isolating critical subsystems from internet exposure, the design enhances overall resilience, addressing contemporary cybersecurity challenges. While theoretical at this stage, this framework aims to inspire further research and development of innovative server security systems..

This research investigates predictive maintenance strategies using the AI4I 2020 dataset, a synthetic dataset reflecting real-world industrial environments. The study identifies failure trends, explores their root causes, and evaluates cost impacts associated with different failure types. Advanced analytics reveal heat dissipation failures (HDF) as the most significant contributor to operational disruptions, accounting for over 74% of total maintenance costs. Through a detailed examination of machine operational parameters and failure distribution patterns, this work proposes tailored maintenance solutions, such as enhanced cooling systems, load optimization, and power stabilizers, to mitigate failures effectively. The findings underscore the importance of prioritizing high-impact failures, leveraging predictive maintenance, and adopting proactive strategies to minimize downtime and optimize costs in industrial settings. This research serves as a comprehensive guide for industries aiming to enhance operational reliability and sustainability.

Large language models (LLMs) have achieved high accuracy in diverse NLP and computer vision tasks due to selfattention mechanisms relying on GEMM and GEMV operations. However, scaling LLMs poses significant computational and energy challenges, particularly for traditional Von-Neumann architectures (CPUs/GPUs), which incur high latency and energy consumption from frequent data movement. While high energy consumption is a crucial challenge for serves, these issues are exacerbated in edge applications with energy-constrained computational resources. DRAM-based near-memory architectures offer high energy efficiency and throughput, but their basic processing elements have limited operations due to stringent area, power, and timing constraints. This work introduces CIDAN-T3D, a novel Processing-in-Memory (PIM) architecture optimized for LLMs. It features an ultra-low-power Neuron Processing Element (NPE) with high compute density (#Operations/Area), enabling efficient in-situ execution of LLM operations by leveraging high parallelism within DRAM. CIDAN-T3D minimizes data transfers, enhances locality, and achieves significant gains in throughput and energy efficiency. Experiments show a 1.3× throughput and 21.9× improvement in energy efficiency compared to existing near-memory designs, making CIDAN-T3D a scalable and efficient solution for LLM-based Gen-AI applications.

This paper presents an automated classification system for identifying abnormalities in Video Capsule Endoscopy (VCE) frames using the DINOv2 vision transformer model. Video capsule endoscopy is a vital non-invasive method for diagnosing gastrointestinal diseases, but manual frame-by-frame analysis can be time-consuming and subject to errors. To address this challenge, we developed a deep learning pipeline capable of classifying ten types of abnormalities, including angioectasia, bleeding, erosion, and polyps. Our model leverages the DINOv2 architecture for feature extraction, combined with fully connected layers for multi-class classification. The training dataset was augmented with random transformations to improve generalization, while test images were processed using standardized resizing and normalization techniques. The model was fine-tuned on over 37,000 labeled VCE frames and evaluated on a separate validation set. Results demonstrated a high overall accuracy of 91%, with particularly strong performance on the "Normal" class (F1-score of 0.97) and a balanced F1-score across other abnormality classes. Despite the dataset imbalance, the DINOv2 model achieved robust results with an average AUC of 0.85. This automated classification approach significantly reduces the need for timeintensive manual analysis and offers a scalable solution for assisting gastroenterologists in early detection of gastrointestinal diseases.

Since the advent of the Internet, information security has become one of the most crucial aspects of information technology and communication. Over the past decade, numerous image steganographic techniques have been developed, primarily concentrating on optimizing payload capacity and image quality. However, a trade-off between these two metrics remains a persistent challenge. This paper presents a multi-layered steganography technique that effectively addresses the limitations of traditional single-layer methods by introducing a novel tiered encryption framework, enabling secure tiered access control. Each message layer is encrypted independently, with unique keys. The resulting ciphertext of each layer is embedded into an image using the least significant bit (LSB) steganography. Experimental results indicate that increasing the layers improves security significantly. This technique is most effective for message sizes up to 1000 KB and when utilizing 3-4 layers. Despite some drawbacks related to image quality, file size, and encoding time, the distortion remains imperceptible, with PSNR values exceeding 60 dB, even in the five-layer configuration. The increased MSE and Chi-Square values demonstrate the method's resilience against steganalysis and brute-force attacks, highlighting its security strengths.

The risk of older drivers' involvement in accidents increases with age, and the damages that they get from crashes are more irreparable than middleaged drivers. The current study uses in-vehicle sensing systems to analyze and classify the driving behaviors of older drivers, precisely those aged 65 and older, using unsupervised machine learning techniques. The study utilizes telematics data from in-vehicle sensors over three years to identify patterns and categorize driving behaviors. The primary objective is to develop a framework that can detect patterns indicative of aggressive or cautious driving styles, leveraging Self-Organizing Maps (SOMs) and Deep Embedded Clustering (DEC) methods to reduce the complexity of the data. The research shows the importance of telematics data in understanding driving behaviors and presents unsupervised learning methods, particularly SOM, as effective methods in visualizing and interpreting complex driving patterns. The results indicate that a 5 × 5 grid SOM and combining DEC with K-means are the most effective methods for determining the optimal number of clusters. The clustering analysis reveals two distinct clusters as the optimal outcome, suggesting that the majority of driving behavior among older drivers in the dataset is marked by cautious driving. The methodologies used are generalizable to other demographics and driving conditions, making the findings relevant for broader applications in traffic safety analysis.

This paper contains a short introduction to Newtonian gravitation. The main focus is on some C++ code.

IntroductionThis paper presents a novel approach to Wildlife Monitoring through non-invasive methods. The idea was originally developed and presented at a hardware hackathon held in 2022 and has not been implemented in practice. Accurate and non-invasive monitoring of wildlife is critical for conserving endangered species and managing ecosystems. Traditional methods, such as GPS collars and radio telemetry, often involve physical handling or tagging, which may cause stress and alter natural behaviors. Recent advancements in IoT and biotechnology present opportunities to revolutionize wildlife monitoring by minimizing human intervention and animal disturbance.This study explores an innovative solution combining IoT sensors, robotic mechanisms, and biotechnology. By leveraging tissue culture to encapsulate robotic systems in a bioengineered exoskeleton, this approach aims to seamlessly integrate technology with the animal’s natural environment, preserving authenticity and improving data accuracy.

Spiking Neural Networks (SNNs) are an essential tool in computational neuroscience, enabling simulations of braininspired models for studying neural dynamics and building intelligent systems. While neuromorphic systems and highperformance computing (HPC) platforms are common choices for SNN simulation, GPUs present a highly competitive alternative by offering superior computational capabilities, better energy efficiency, and broader flexibility. This paper explores how GPUs can outperform traditional approaches in simulating large-scale SNNs, discussing their role in overcoming memory bandwidth limitations, enabling real-time applications, and supporting computational steering for interactive simulations. The advantages of using GPUs in neurorobotics and the potential is highlighted for extending their application to multi-GPU systems for larger models.

In this work, we propose a novel approach for the detection of chipless radio frequency identification (RFID) signals. The method is founded upon the application of transformations to the measurement, in conjunction with the utilization of Artificial Intelligence (AI) algorithms. In the initial stage of the process, frequency-related features are measured. Subsequently, a time-frequency representation of these measurements is generated through the application of the Inverse Fourier Transform (IFT), a time-gating strategy, and the Continuous Wavelet Transform (CWT). The resulting representation is then employed as input to a shallow convolutional neural network (CNN), which is able to learn complex patterns while being able to generalize to new measurements. Furthermore, the proposed scheme incorporates a filtering process, based on the probabilities derived from the model, to filter out low-confidence predictions. To assess the performance of the proposed method, we considered a population of 16 tags. We collected 4,800 measurements for the training phase and 2,400 measurements for the testing phase in a real environment, collected at different days, and within a distance range of 50-140 cm from the tag to the antenna. The proposed method exhibited accuracies of 94% in the 110-140 cm range, 99% in the 80-110 cm range, and 100% in the 50-80 cm range, showcasing its suitability for chipless RFID detection.

This paper introduces a new receiver coil configuration that dramatically enhances energy transfer efficiency in Dynamic Wireless Power Transfer (DWPT) systems. By reimagining the traditional coil layout, we propose the Vertical Coil (VC) design, which leverages multiple vertically aligned coils wound around a ferrite core to capture magnetic flux more effectively than the conventional DDQ horizontally oriented receiver coils. Full scale VC and DDQ models have been studied by simulation and small, bench-models have been realized and studied both experimentally and by simulations. The VC configuration not only results in a higher peak efficiency when the receiver is right above the transmitter, it also spans a larger range of efficient energy transfer. Our findings demonstrate that the VC receiver has the potential to offer better performances and higher energy transfer efficiency, thereby contributing to the development of dynamic wireless power transfer technology and the EV industry.