Search

In this paper, we address a class of distributed robust cooperative fault-tolerant control problems for discrete-time multi-agent systems (MASs). The discrete-time MASs considered in this paper have both uncertainties and actuator faults, and a distributed intermediate variable estimator that does not require a fault estimation matching condition is designed for obtaining the estimation information of the actuator faults. In addition, to further obtain higher design freedom and better estimation performance, the intermediate variable estimator considers both centralised and distributed output estimation errors as feedback terms. A fault-tolerant control protocol is designed using the fault estimation information to ensure the reliable operation of discrete-time MASs. It is worth noting that in the method proposed in this paper, there is no need to know the boundary information about the faults and their changing rates, and the method is applicable to discrete-time MASs with directed communication topology. In order to ensure the solvability of the Linear Matrix Inequality (LMI) for better obtaining of the gain matrices of the designed fault estimators, further decoupling and dimensionality reduction of the LMI are carried out in this paper. Finally, a numerical simulation and a network of four one-link flexible joint manipulator systems simulation verify the truthfulness and effectiveness of the method proposed in this paper.

Early fault warning for large-scale high-speed rotating machinery can effectively reduce unplanned downtime and avoid major safety accidents. Aiming at the problems of difficult screening of multi-source common sensitive features, the challenging training of neural networks with a small number of sensitive features, and the difficulty of directly using generative adversarial networks for early fault warning, this paper constructs an early fault warning model based on multi-source common sensitive features and an improved Wasserstein generative adversarial network, proposing an early fault warning method for rotating machinery. The model was verified by using the open XJTU-SY bearing laboratory data, the P3409A centrifugal pump bearing fault engineering case data of a petrochemical company and the rotor system engineering case data of a circulating hydrogen centrifugal compressor of a petrochemical company. The early fault warning method of rotating machinery proposed in this paper warns the bearing fault of centrifugal pump 160 hours in advance and the rotor system fault of centrifugal compressor 1330 minutes in advance. Compared with the two published methods, the proposed method has better early fault warning effect, better normal and abnormal health index discrimination and less false warning.

Existing artificial intelligence (AI) models for diagnosing knee osteoarthritis (OA) have faced criticism for their lack of transparency and interpretability, despite achieving medical-expert-like performance. This opacity makes them challenging to trust in clinical practice. Recently, explainable artificial intelligence (XAI) has emerged as a specialized technique that can provide confidence in the model’s prediction by revealing how the prediction is derived, thus promoting the use of AI systems in healthcare. This paper presents the first survey of XAI techniques used for knee OA diagnosis. This survey identified 78 AI-based primary knee OA diagnostic test accuracy studies, of which 70 (89.7%) employed XAI. In 34 out of 70 (48.6%) of studies, XAI was utilized for the goal of visualization of predictions. Gradient-weighted class activation mapping (GradCAM) is the most common technique, being used in 24 out of 70 studies (34.3%), followed by SHapley Additive exPlanations (SHAP), being used in 9 out of 70 (12.9%) studies. All included studies analyzed the outcomes generated by XAI methods through qualitative analysis. However, only three studies utilized quantitative measures to evaluate the reliability of the XAI outcomes. We also observed that 64.3% of the studies utilized widely-circulated dataset, namely Osteoarthritis Initiative (OAI) extensively.The XAI techniques are discussed from two perspectives: data interpretability and model interpretability. Our paper provides an overview of XAI’s potential towards a more reliable knee OA diagnosis approach and helps to encourage its adoption in clinical practice.

Augmented Reality (AR) has experienced a significant resurgence in popularity and interest in recent years. Despite numerous surveys and reviews in the field, information remains scattered and challenging to consolidate for newcomers. This paper addresses this gap with a comprehensive study of AR’s state of the art. We begin with an introduction to AR and related terminologies. We then describe three enabling of AR and four enhancing technologies. A survey then covers different types of commercial AR hardware with detailed specifications and elaborations. We then address AR software and discuss its basic and advanced features, mapping them to software with feature matrices. Based on a thematic literature analysis, we extract a comprehensive set of guidelines for developing AR solutions into seven themes, from ideation to best practices for implementation and deployment. Finally, we explore some essential challenges in the field of AR. This paper serves as an essential resource for those looking to understand what AR is, what is needed to get started, how to approach development, and what the future holds for AR. By consolidating essential information and providing a solid foundation, this paper aims to help researchers and developers capitalize on the current cycle of interest in AR.

This paper introduces three new variations of the HFO-1 (Honey Formation Optimization with Single Component) algorithm, namely HFO-1a, HFO-1b and HFO-1c, adapted to address the optimal power flow (OPF) problem. The original HFO-1 algorithm has shown success in solving various numerical problems in recent years; however, it assumes a single search range for all dimensions of the solution space, making it unsuitable for direct application to the OPF problem. Modifications to both the honey formation and mixing phases of the HFO-1 algorithm were made to improve solution quality and convergence speed, resulting in three new variants of HFO-1. The newly developed variants aim to minimize even the most challenging objective functions of the complex OPF problem, which has been further complicated by the integration of renewable energy sources into power systems. The paper provides a comprehensive and transparent comparison of the three types of IEEE 30-bus test systems and 118-bus test systems with existing methods, meticulously adhering to practical, technical, operational, and safety constraints. Following successful results on the CEC 2021 standard benchmark functions, the proposed HFO-1 variants have been thoroughly validated through extensive analysis. Experimental results demonstrate that the proposed approach can achieve lower costs (800.5972/hour and 800.3871/hour) in two types of IEEE 30-bus systems without integrating renewable resources while maintaining system constraints. Furthermore, HFO-1a (achieving 3.0776261 MW) and HFO-1b achieve the lowest values in the literature with a multi-fuel cost of (646.375893 /h) and a valve point effective fuel cost of (823.981360 /h), respectively, while HFO-1c exhibits a voltage deviation of (0.083498 p.u.) and Prohibited Operating Zones (POZ) cost of generator (800.665078 /h).

As technology advances, the use of Machine Learning (ML) in cybersecurity is becoming increasingly crucial to tackle the growing complexity of cyber threats. While traditional ML models can enhance cybersecurity, their high energy and resource demands limit their applications, leading to the emergence of Tiny Machine Learning (TinyML) as a more suitable solution for resource-constrained environments. TinyML is widely applied in areas such as smart homes, healthcare, and industrial automation. TinyML focuses on optimizing ML algorithms for small, low-power devices, enabling intelligent data processing directly on edge devices. This paper provides a comprehensive review of common challenges of TinyML techniques, such as power consumption, limited memory, and computational constraints; it also explores potential solutions to these challenges, such as energy harvesting, computational optimization techniques, and transfer learning for privacy preservation. On the other hand, this paper discusses TinyML’s applications in advancing cybersecurity for Electric Vehicle Charging Infrastructures (EVCIs) as a representative use case. It presents an experimental case study that enhances cybersecurity in EVCI using TinyML, evaluated against traditional ML in terms of reduced delay and memory usage, with a slight trade-off in accuracy. Additionally, the study includes a practical setup using the ESP32 microcontroller in the PlatformIO environment, which provides a hands-on assessment of TinyML’s application in cybersecurity for EVCI.

Given the substantial role of electric vehicles (EVs) in decarbonization, the demand for EVs has become widespread. However, batteries remain a major impediment, mainly due to critical environmental concerns. Inductive Power Transfer (IPT) is an effective technique to eliminate the reliance on batteries by enabling charging without physical connection. This paper conducts a thorough analytical review of the most recent advancements in the design of IPT couplers and identifies current gaps and future opportunities. Different aspects of IPT systems are reviewed, various coil designs are compared, and their performance is analyzed. In-depth theoretical analysis is conducted through mathematical modeling, and the electric and magnetic equations are derived to ensure accurate outcomes. Moreover, a magnetic field calculation tool using a Python script has been developed to perform the magnetic computations and produce 3D graphs of the total magnetic flux linkage. Additionally, performance evaluations have been conducted for complex couplers based on simulation results executed in Ansys Electronics software. To ensure comprehensiveness, the trade-off trends in the design of magnetic couplers are presented, and relevant optimization techniques are reviewed and explained. This paper is anticipated to be a comprehensive reference for researchers seeking both in-depth theoretical analysis of IPT systems and a thorough review of coupler designs.

In this paper, we present hardware implementations of the lightweight TinyJAMBU cipher with reduced power consumption using a mechanism based on shift register parallelization. The power consumption in digital circuits depends linearly on the switching activity of the logic gates. The parallelization technique reduces the number of switches per clock cycle of the shift registers, which can significantly reduce power consumption. This technique has been applied to the TinyJAMBU cipher, a finalist in the NIST lightweight cryptography standardization process with the lowest resource and power consumption. The implementations we present use the logical parallelization technique in the cipher’s NLFSR, which is the basic block of TinyJAMBU, and in the key register. Measurements with data from various post-place and route simulations, performed for the Xilinx 7-Series family of FPGAs, are presented to demonstrate the effectiveness of the proposed technique in reducing power consumption, achieving a reduction of more than 30% in dynamic power consumption compared to the standard implementation, with almost no increase in resource consumption. Therefore, the ciphers hardware implementations proposed in this paper are highly suitable for use as hardware peripherals in applications with severe constraints on available resources and power, such as low-power microcontrollers that are widely used in the IoT field.

The use of LiDAR technologies offer an effective alternative for enhancing monitoring and surveillance tasks in secured areas, or hard-to-access areas. Conventional solutions allow detection and identification capabilities, which are valuable for object and pattern recognition in environments with special conditions. This paper proposes a method for structure a dataset by collecting data from LiDAR sensors and reconstructing 3D point cloud scenes into images. Furthermore, training with these data using semantic segmentation algorithms allows for the detecting of people in monitoring and surveillance contexts. The process involves data acquisition techniques, preprocessing, and the application of tools and technologies for semantic segmentation in the post-processing stage. Data is acquired from RPLidar sensor with a spatial resolution of less than 0.4 degrees per step, ensuring greater accuracy in detecting small objects. Scanning distances between 0.15 and 12 meters allow for the capture of 941 points per rotation angle. The dataset is complemented with labels and images containing binary masks of regions of interest for people detection using semantic segmentation. The main contribution of this paper is the methodological description of constructing a dataset comprising 1,254 images in various scenarios, including indoors and outdoors, with different levels of occlusion and illumination. This dataset facilitates the development of machine learning algorithms and deep learning algorithms for pattern recognition, such as Mask R-CNN (89.9% mean class accuracy) and U-NET (90.5% mean class accuracy). The results demonstrated an approximate 2% improvement in mean class accuracy compared to other state-of-the-art algorithms in semantic segmentation for point clouds.

The transmission system of a tractor is a crucial component, so it is crucial to promptly identify and correctly diagnose faults in it. However, due to the limited samples of faults occurring during its operational processes, employing existing fault diagnosis methods directly yields unsatisfactory results. This paper proposes a fault diagnosis model combining Time Generative Adversarial Networks (Time GAN) and Transformer. To enhance diagnostic accuracy, we first employ Time GAN for data augmentation, addressing the issue of imbalanced fault samples in practical scenarios. Then, we integrate a Transformer network with improved multi-head self-attention mechanisms, leveraging the advantages of the Transformer’s encoder-decoder architecture and attention mechanism to enhance diagnostic performance. Bearing data from Case Western Reserve University (CWRU) was used to validate the diagnostic performance of the proposed model, while gear data from an experimental rig built by the author was used to validate the model’s generalization capability. Experimental results indicate that the accuracy reached 98.96% and 95.36% in CWRU Dataset and Self-made Dataset respectively. In strong noise environments, the accuracy remains above 93%. In conclusion, the diagnostic model presented in this paper can reliably diagnose tractor transmission system problems in few-sample conditions and noise environments compared to traditional machine learning models.

Deep reinforcement learning (DRL) has emerged as a promising technique for optimizing the deployment of unmanned aerial vehicles (UAVs) for data collection in wireless sensor networks (WSNs) and Internet of Things (IoT) applications. With DRL, UAV trajectory can be optimized, optimal data collection points can be determined, sensor node transmissions can be scheduled efficiently, and irregular traffic patterns can be learned effectively. In view of the significance of DRL for UAV-assisted IoT research in general and, more specifically, its use for time-critical applications, this paper presents a review of the existing literature on UAV-aided data collection for WSN and IoT applications related to the application of DRL to minimize the Age of Information (AoI), a recent metric used to measure the degree of freshness of transmitted information collected in data-gathering applications. This review aims to provide insights into the state-of-the-art techniques, challenges, and opportunities in this domain through an extensive analysis of a sizable range of related research papers in this domain. It discusses application areas of UAV-assisted IoT, such as environmental monitoring, infrastructure inspection, and disaster response. Then, the paper focuses on the proposed works, their optimization objectives, architectures, simulation libraries and complexities of the various DRL-based approaches used. Thereafter discussion, challenges, and some opportunities for future work are provided. The findings of this review serve as a valuable resource for researchers and practitioners, guiding further advancements and innovations in the field of DRL for UAV-aided data collection in WSN and IoT applications.

The high-dimensional data fitting modeling based on hybrid neural network models has been a hot topic in data mining research in recent years. However, due to the curse of dimensionality and limited number of trainable samples, the model’s training accuracy cannot continue to improve after reaching a certain level. In this paper, we propose a hybrid neural network model with a dual structure to further improve the training accuracy of the basic model by controlling the data dimension and adjusting the model structure. The proposed model improves the ratio of sample quantity to dimension, enhances the stability and generalization ability of the model, and shortens the training time. The test results on multiple datasets demonstrate that the comprehensive accuracy of the proposed model is 13% higher than that of the basic single-branch model. Although the current Transformer model surpasses the proposed DmHybNNs model in terms of accuracy, its model complexity is significantly higher than that of the DmHybNNs model. The low-complexity DmHybNNs model is more suitable for deployment on low-power and low-computing-power platforms. This research paper is both theoretical and practical, providing some new ideas and methods for modeling high-dimensional nonlinear systems.

Model identification errors and time delays are often presented in power systems, which will deteriorate the effect of power system wide-area damping control. To solve the problems of random time delays and model parameter identification errors on system damping control, this paper firstly uses the pseudo-random noise excitation as the identification signal and establishes a closed-loop model for the power system under small interference conditions. Secondly, a Vinnicombe-Adaptive Multi-parameter Differential Evolutionary algorithm (VAMDE) proposed in this paper takes into account of the algorithm model identification error, directly predicts the value of the random time delay parameter, and gives the optimal damping controller that can effectively suppress low-frequency oscillations in power system. The most important thing is that the random time delay is compensated by the improved Smith Predictor. Finally, a four-machine and two-area system is tested. Compared with traditional iterative identification method, simulation results show that the proposed method can effectively improve the damping control problem caused by the model identification error and communication time delay.

Tattoos can serve as an essential source of biometric information for public security, aiding in identifying suspects and victims. In order to automate tattoo classification, tasks like classification require more detailed image content analysis, such as semantic segmentation. However, a dataset with appropriate semantic segmentation annotations is currently lacking. Also, there are countless ways to categorize tattoo classes, and many are not directly categorizable, either because they belong to a specific artistic trait or characterize an object with previously undefined semantics. An effective way to overcome these limitations is to build recognition systems based on open-set assumptions. Nevertheless, state-of-the-art open set approaches are not directly applicable in tattoo semantic segmentation, mainly due to the significant class imbalance (predominant background). To the best of our knowledge, this paper is the first to explore semantic segmentation in closed and open-set scenarios for tattoos. In this sense, this paper presents two key contributions: (i) a novel large-margin loss function and generalized open-set classifier approach and (ii) an open-set tattoo semantic segmentation dataset with a publicly accessible test set, enabling comparisons and future research in this area. The proposed approach outperforms other methods, achieving 0.8013 of AUROC, 0.6318 of Macro F1, 0.4900 of mIoU, and notably 0.2753 of IoU for the unknown class, demonstrating the feasibility of this approach for automatic tattoo analysis. The paper also highlights key limitations and open research areas in this challenging field. Dataset and codes are available at https://github.com/Brilhador/tssd2023.

Under-resourced automatic speech recognition (ASR) has become an active field of research and has experienced significant progress during the past decade. However, the performance of under-resourced ASR trained by existing methods is still far inferior to high-resourced ASR for practical applications. In this paper, speech data from languages that share the most phonemes with the under-resourced language are selected as supplementary resources for meta-training based on the Model-Agnostic Meta-Learning (MAML) strategy. Besides supplementary language selection, this paper proposes a dynamic sampling method instead of the original random sampling method to select support and query sets for each task in MAML to improve meta-training performance. In this study, Taiwanese is selected as the under-resourced language, and the speech corpus of five languages, including Mandarin, English, Japanese, Cantonese, and Thai, are chosen as supplementary training data for acoustic model training. The proposed dynamic sampling approach uses phonemes, pronunciation, and speech recognition models as the basis to determine the proportion of each supplementary language to select helpful utterances for MAML. For evaluation, with the selected utterances from each supplementary language for meta-training, we obtained a Word Error Rate of 20.24% and a Syllable Error Rate of 8.35% for Taiwanese ASR, which were better than the baseline model (26.18% and 13.99%) using only the Taiwanese corpus and other methods.

The end of Moore’s Law and Dennard Scaling has slowed processor improvements in the past decade. While multi-core processors have improved performance, they are limited by the application’s level of parallelism, as prescribed by Amdahl’s Law. This has led to the emergence of domain-specific processors that specialize in a narrow range of functions. Smart Network Interface Cards (SmartNICs) can be seen as a revolutionary technology that combines heterogeneous domain-specific processors and general-purpose cores to offload infrastructure tasks. Despite the impressive advantages of SmartNICs and their importance in modern networks, the literature has been missing a comprehensive survey. To this end, this paper provides a background encompassing an overview of the evolution of NICs from basic to SmartNICs, describing their architectures, development environments, and advantages over legacy NICs. The paper then presents a comprehensive taxonomy of applications offloaded to SmartNICs, covering network, security, storage, and compute functions. Challenges associated with SmartNIC development and deployment are discussed, along with current initiatives and open research issues.

The paper presents a novel framework for ship International Maritime Organization (IMO) number Identification using unmanned aerial vehicles (UAVs). It comprises three integrated modules: ship IMO region detection, text detection within detected IMO regions, and IMO number extraction from detected text. Furthermore, considering real-world implementation, the paper details the framework’s practical deployment on UAV and proposes an algorithm for efficiently extracting IMO numbers from the detected text. To ensure robust performance evaluation, a comprehensive evaluation metric is established for screening various ship IMO region detection, text detection, and recognition algorithms. Through extensive experimentation, YOLOx_S, DB_R18, and SVTR were identified as optimal for ship IMO region detection, text detection, and text recognition respectively. Finally, we acknowledge the presence of potential false detections in the results and emphasize that while the comprehensive evaluation metric offers valuable insights, it should not be the sole criterion for algorithm selection.

Truck platooning uses networking technology and automated driving support systems to join multiple trucks in a group. When these vehicles interact for particular journey stages, such as on highways, they autonomously maintain a predefined, tight spacing among themselves. Platooning improves transportation by making better use of highways, delivering cargo faster, and minimizing congestion in traffic. Therefore, safety and platooning are two important attributes of an intelligent truck system. This paper discusses a quantum-safe blockchain-empowered authentication mechanism for autonomous truck platooning. The proposed idea is to use blockchain to combine multiple nodes and ensure authenticity with the help of an aggregation technique. The proposed design ensures authenticity to the system due to hard assumptions:1) Learning With Error (LWE) and 2) Short Integer Solution (SIS) on random module generated lattice. This paper uses operations over module lattices to be more efficient than general lattices. We can perform operations on module lattices with the help of fast algorithms for polynomial arithmetic.

Due to the sheer volume and variety of data in industrial manufacturing, manually creating features can be costly. Therefore, appropriate feature processing methods become crucial. Most existing feature processing methods abstract feature engineering as a feature search problem, i.e., finding feature transformations that optimize model performance. However, for automated feature engineering, the number of searches and the number of transformation combinations are huge. Therefore, we use a factorization-based model that measures interactions in terms of vector products, automatically learns patterns of combined features, and generalizes them to unseen features. Prior to this paper, the DeepFM algorithm (which combines an FM model with a deep neural network model) and the xDeepFM algorithm (which proposes a novel Compressed Interaction Network (CIN) designed to make feature interactions explicit) were available. The LRCIN proposed in this paper focuses on improving the CIN network in the xDeepFM method, by introducing a low-rank approximation method in the CIN network to reduce the number of parameters, and adding an attention mechanism after the CIN to ensure the accuracy of the model. The experimental results show that our method can effectively reduce the time complexity of the model and improve the model accuracy to some extent.

Mobility management is a critical and challenging requirement for 5G networks and their successors, as numerous highly mobile user equipment (UE), including massive Internet-of-Things (IoT) devices, generate extreme traffic volumes. In the integration of non-terrestrial networks (NTNs) with 5G, mobility management is of utmost importance to track these highly mobile UE/IoT throughout the network coverage area for the purpose of data-packet delivery. In this context, we thoroughly discuss the integration of satellite-based NTNs with 5G technology, which has become increasingly active in recent years, and studies are still ongoing to develop innovative frameworks to provide seamless connections to the extremely increasing density of high-mobility UE/IoT. The goal of this integration is to extend coverage and improve reliability beyond current 5G networks, enabling ubiquitous connectivity in remote areas and delivering high-speed data to UE/IoT. However, this integration poses numerous challenges in terms of mobility management, especially concerning the power constraints and signaling overhead of UE/IoT, as these devices are limited in battery power and processing capabilities. Handling these challenges is crucial to ensure successful integration and provide seamless connectivity to UE/IoT. In this paper, we highlight recent research efforts and potential solutions in these areas, contributing to the ongoing development of non-terrestrial 5G networks and improving global connectivity. To the best of our knowledge, this paper is the first study to emphasize and critically assess different mobility management solutions for satellite-based 5G NTNs, evaluating their impact on UE/IoT battery power consumption and signaling overhead, with implications extending to 6G networks.

With the widespread availability and utilization of 3D meshes across various applications, the need for accurately assessing their visual quality has become increasingly important. Despite the significance of this task, the literature offers few No-Reference (NR) approaches for evaluating the visual quality of 3D meshes. In response to this gap, this paper proposes a novel NR approach tailored specifically to score the quality of 3D meshes. After rendering a 3D mesh into 2D views and patches, a pre-trained convolutional neural network automatically extracts deep features from. These features are then employed in a Multi-Layer Perceptron regressor to predict the quality score of the rendered images. The obtained scores are combined with their corresponding BRISQUE scores, and an additional MLP regressor is used to predict the final score. We present experimental results demonstrating the effectiveness and robustness of our approach across a diverse range of 3D mesh datasets. Comparative analyses with existing NR methods underscore the superior performance and versatility of the proposed approach. Overall, this paper contributes to the advancement of NR techniques for assessing 3D mesh quality, offering a valuable tool for researchers, practitioners, and developers working with 3D models across various domains.

Recently, the combination of convolutional neural network (CNN) and long short-term memory (LSTM) exhibits better performance than single network architecture. Most of these studies connect LSTM networks behind CNNs. When operating on hardware, the current design of CNN-LSTM is similar to a pipeline architecture. However, the classic structure lead to a feature loss when data is sent to LSTM since CNN is not good at extracting temporal features. At the same time, as the depth and scale increases, it will bring a huge amount of computation, which makes hardware implementation difficult. Based on that, a parallel CNN-LSTM architecture is proposed, in which two networks extract features from the input data synchronously, being proven to be more effective than classical CNN-LSTM. This paper designs a parallel CNN-LSTM computing device based on FPGA. The device is divided into control unit and operation unit. Control stream and data stream transport between the two units, ensuring the proper running of the device. A highly parallel multi-channel convolution layer and pooling layer are designed to improve the calculation efficiency. A 4-stage pipeline structure is adopted to implement the LSTM part. This paper makes full use of on-chip BRAM to design a look-up table for activation function approximation, reducing the resource consumption by 95% compared with the traditional polynomial approximation. Finally, we verify our device under cooperative spectrum sensing (CSS) and handwritten classification scenarios. Proposed device reaches higher accuracy in two scenarios compared with classic CNN-LSTM structure as well as faster calculating speed, and the overall project power is limited below 2W. The scalability and limitation of this computing device are also discussed.

In various fields, such as robotics navigation, autonomous driving, and augmented reality, the demand for the reconstructing of three-dimensional (3D) scenes from two-dimensional (2D) images captured by a camera is growing. With advancements in deep learning, monocular depth prediction research has gained momentum, leading to the exploration of 3D reconstruction from a single image. While previous studies have attempted to restore occluded regions by training deep networks on high-resolution 3D data or with jointly learned 3D segmentation, achieving perfect restoration of occluded objects remains challenging. Such 3D mesh generation methods often result in unrealistic interactions with graphic objects, limiting their applicability. To address this, this paper introduces an amodal depth estimation approach to enhance the completeness of 3D reconstruction. By utilizing amodal masks that recover occluded regions, the method predicts the depths of obscured areas. Employing an iterative amodal depth estimation framework allows this approach to work even with scenes containing deep occlusions. Incorporating a spatially-adaptive normalization (SPADE) fusion block within the amodal depth estimation model effectively combines amodal mask features and image features to improve the accuracy of depth estimation for occluded regions. The proposed system exhibits superior performance on occluded region depth estimation tasks compared to conventional depth inpainting networks. Unlike models that explicitly rely on multiple RGB or depth images to handle instances of occlusion, the proposed model implicitly extracts amodal depth information from a single image. Consequently, it significantly enhances the quality of 3D reconstruction even when single images serve as input. The code and data used in the paper are available at https://github.com/Seonguke/Occlusion-aware-Amodal-Depth-Estimation-for-Enhancing-3D-Reconstruction-from-a-Single-Image/ for further research and feedback.

Precise forecasting of solar power output is crucial for integrating renewable energy into power networks, improving efficiency and dependability. This study assesses the efficacy of several Machine Learning (ML) algorithms in predicting solar power generation through a detailed performance comparison. This paper analyzes six algorithms: CatBoost, Gradient Boosting Machines (GBMs), Multilayer Perceptron (MLP) regressor, Support Vector Machines (SVMs), XGBoost, and Random Forest (RF). Using a dataset of 4213 sets of solar power generation data, each model was trained and tested, with performance evaluated based on R-squared (R2) scores for the whole dataset, training set, and test set. Also, this study examined the mean and standard deviation of test set predictions to gauge how consistent each model was. The results showed that RF had the highest overall R2 score of 0.940 and a training set score of 0.971. XGBoost demonstrated exceptional performance on the test set, attaining a high R2 score of 0.822. CatBoost and GBMs exhibited strong performance, albeit with slightly lower R2 values of 0.786 and 0.829, respectively. Although the MLP regressor and SVMs exhibited high training scores, they encountered difficulties in generalizing to unfamiliar data. This paper highlights the effectiveness of combining XGBoost and RF techniques in improving the accuracy of solar power forecasts. The investigation focuses on enhancing the precision and reliability of renewable energy projections through a comprehensive comparison of various contemporary ML techniques.

In this paper, we introduce an innovative generalized Lyapunov theorem and a novel bounded real lemma designed for continuous-time linear singular systems with Caputo fractional derivative of order $\alpha $ , with the constraint $1 \leq {\alpha }\lt 2$ . We initially present a condition that is both necessary and sufficient for establishing the admissibility of singular fractional-order systems (SFOSs). This condition is articulated through strict linear matrix inequalities (LMIs). Following this, we demonstrate that a SFOS satisfies ${H_{\infty }}-$ norm requirement if and only if two strict LMIs are feasible. The key advantage of the presented LMI conditions is that only one matrix variable needs to be solved. Ultimately, this paper concludes by presenting illustrative examples that highlight the practical effectiveness of our theoretical findings.

Urban day-to-day travel system is a complex nonlinear system, its current status depends not only on the status of the whole urban transportation system at the previous point in time, but also on the travel decision made by each traveler. Moreover, information environment generated by various intelligent terminals like smart phones, has even increased the complexity of urban day-to-day travel system, for micro level route choice decision made by each traveler can be affected by macro level road traffic flow information at any time, and micro route choice behavior can generate new macro traffic information. Nowadays, there is a lack of macro-micro integrated models to explain urban day-to-day travel system, especially its development in information environment. In this paper, complex system dynamics has been proposed and used to explain urban day-to-day travel system in a macro-micro integrated way. The main contribution of this paper is twofold: in micro level, each traveler in urban day-to-day travel system in information environment is endowed with unique physical properties, and the time correlation effect of individual travelers is described; in macro level, the traffic flow fluctuations under the influence of individual micro decisions, the stylized facts of urban day-to-day travel has been deduced, and an effective description of urban day-to-day travel phenomena has been formed.

This paper investigates the performance and parametric design of a new foundry-based silicon field-effect transistor (FET) sensing platform known as the open-gate junction field-effect transistor (OG-JFET) for bio/chemical sensing applications. The fabrication process of the OG-JFET relies on a standard foundry process, requiring the establishment of parametric design rules to understand the effect of crucial sensor performance factors, including transconductance (g $_{\mathrm {m}}$ ) and device efficiency ( $\eta =$ gm/I $_{\mathrm {ds}}$ ). The study examines the impact of various geometric parameters (e.g., channel length and thickness) and material-related parameters (such as boron and phosphorous impurity doping levels) on sensor performance. Simulations provide insights and guidelines for the efficient design and characterization of the OG-JFET, focusing on enhancing gm and maximizing $\eta $ for biosensing applications. Experimental measurements of the OG-JFET demonstrate a current range of $\sim ~200~\mu $ A/ $\mu $ m, a high gm of approximately $\sim ~1700~\mu $ S ( $340~\mu $ S/ $\mu $ m), and $\eta $ of $\sim ~4.5$ V−1 (for a channel length of $5~\mu $ m), which are of importance for circuit design and biosensing application of OG-JFET. These results showcase the performance of this sensor compared to other silicon-based FET platforms for internal signal amplification in sensing applications of OG-JFET. The findings of this paper offer valuable guidelines for the design of sensors based on OG-JFET technology, enabling gaining insights into the impact of sensor structure on sensing performance. Also, we have demonstrated how two performance parameters can be utilized to compare two different designs of OG-JFET, which is useful for future designers.

As the economy and society continue to develop, the types of underwater targets are becoming increasingly diverse, accompanied by a corresponding increase in environmental noise. This noise presents a significant challenge for the recognition of underwater target radiation noise, as it can result in the loss of effective information during the noise processing stage. This, in turn, reduces the accuracy of underwater target radiation noise recognition. This paper proposes an underwater target radiation noise recognition method based on the enhanced image method. The method transforms the original underwater target radiation noise signal into an enhanced image, builds a convolutional neural network with the enhanced image as an input, and uses the convolutional neural network’s ability to classify images to recognise and classify underwater target radiation noise. The experimental results demonstrate that the training time of the method described in this paper is longer than that of the traditional machine learning method, yet the recognition and classification accuracy is significantly higher.

This study investigates the application of machine learning (ML) algorithms to enhance the precision of wine quality assessment, focusing specifically on Portuguese red wine. Amidst the growing interest in leveraging artificial intelligence (AI) for sensory analysis, our research distinguishes itself by employing a rigorous methodological framework. Our approach, named the ‘Incremental Analysis of Baseline Accuracy,’ identifies the chemical variables most predictive of wine quality. This framework aims to streamline the predictive process by pinpointing key variables that significantly influence quality assessments. In this paper, we demonstrate the feasibility of a methodology that precisely determines the criticality of chemical inputs, both their exact values and their correct order, to identify which inputs significantly contribute to the quality assessment of a sensory perception, such as taste. The centerpiece of our paper is a vibrant 3D pie chart that illustrates the percentage criticality of different input variables for perceiving the quality of red wine. This chart symbolizes the essence of our paper: a ‘pie’ representing the empirical conclusion, not mere conjecture. Through this paper, we have shown that it is possible to quantify a qualitative, perceptual aspect like taste perception, which is often believed to be assessable only through subjective conjecture. Moreover, our findings, facilitated by the Incremental Analysis of the Baseline Accuracy method, demonstrate that this perception can be systematically quantified, challenging traditional assumptions about sensory analysis.

After end of decades of loadshedding, the power system of Nepal is moving towards the scenario of self-sufficiency. The Government of Nepal has formulated a policy to add 20 GW of hydropower in the next ten years, making a ten-fold increase compared to the current capacity. Therefore, there is a high need for the assessment of the power system stability of the current Nepal grid to mitigate any potential issues that can occur due to the addition of this huge generation. This paper conducts a comprehensive assessment of the power system stability namely voltage (static), angle (transient and small signal), and frequency stability. The proposed assessment methodology consists of three stages: the first stage constitutes the selection and modeling of the components (generator, load, line, excitation system, governor, renewable energy resources) of the power system of Nepal. The second stage then uses the stability assessment tools: time domain simulations for frequency and transient stability, modal (eigenvalue) analysis for small signal stability and voltage stability, and sensitivity analysis for identification of critical areas prone to instability. The result shows that the current grid of Nepal has a low margin of the small signal stability. The critical clearing time of the system is approximately 0.51s. The result also show that there are five potential voltage instability areas in Nepal but is robust in terms of the frequency stability. Additionally, the integration of photovoltaic generation to the Nepal grid has also been studied in this paper. The study shows that the margin of the low frequency oscillatory stability decreases with the penetration of photovoltaic generation of the power grid of Nepal.

Pavement distress detection is crucial in road health assessment and monitoring. However, there are still some challenges in extracting pavement distress based on deep learning: such as insufficient segmentation, extraction errors and discontinuities. In this paper, we propose DARNet, a network for pavement distress extraction. A Distress Aware Attention Module (DAAM) is proposed to solve the problem of discontinuity in distress extraction due to inaccurate recovery of distress pixels during upsampling. Based on the characteristics of distress morphology, a Refinement Extraction Module (REM) is designed to effectively capture horizontal and vertical pavement damage features by fusing high-level and low-level features, which improves the accuracy of the model in extracting details of pavement damage information. Finally, a Weighted Cross-Entropy Loss function (WCEL) is introduced to assign weights according to the distance of the pixel point to the boundary of the distress, which solves the problem that the traditional cross entropy function treats each pixel point equally. We also propose a set of pavement distress datasets LNTU_RDD_GS, and the experimental results show that DARNet can reach 82.68% mIoU and 90.13% F score in the datasets in this paper, 80.63% mIoU and 88.35% F score in the four public datasets.

Improving the accuracy of the six-axis force sensor Six-axis force sensors (hereafter collectively referred to as 6a-sensors) is a systematic problem, which mainly involves the optimization and improvement of three aspects: the design of the elastic body of the 6a-sensors, the acquisition and processing of the weak signal, and the decoupling of the inter-dimensional data of the force and torque of each dimension. To attain high stiffness, sensitivity, and minimal coupling, a design scheme of cross beam elastomer based on titanium alloy material is proposed. In order to realize the acquisition and processing of multi-channel weak differential signals with high gain, high linearity and precision, a high-speed programmable weak signal processing circuit design scheme is proposed. To efficiently address the interdependence issue among signal dimensions in the 6a-sensors. This paper innovatively proposes a BP-PSO decoupling algorithm using Particle Swarm Approach (PSO) to optimise the optimiser of BP neural network. To validate the algorithm’s effectiveness, comparative experiments are conducted with the 6a-sensors. designed and studied in this paper by sampling and self-designed calibration system. The sensitivity of the 6a-sensor designed in this paper reaches 4mV/V. In terms of accuracy, there is an improvement of about 1%. In terms of crosstalk, there is an improvement of about 0.2%. providing a new and improved idea for the six-dimensional force decoupling algorithm. In order to further verify the practicability of the algorithm, a flexible surface grinding robot based on six-axis force is developed based on Siasun robot platform. The grinding strength control is accurate, and the running trajectory is uniform and smooth, which can effectively meet the practical application requirements of flexible and intelligent control.

Autonomous vehicle lane change maneuvers are an intricate task that requires multiple subsystems to work together to ensure the maneuver’s safety and efficiency. This research paper investigates lane change maneuvers of autonomous vehicles, utilizing a trajectory generated using a sinusoidal function considering the ISO 3888 standards. The driving maneuver steering angle input is derived from the generated trajectory, which is adjusted using model predictive control (MPC) to calculate an optimal steering angle for lateral movement and manage throttle input to maintain longitudinal stability. Throughout lane change maneuvers, the vehicle’s states are estimated using model-based Kalman filters, relying on input and measurement data from inertial sensors. The paper compares four state estimator filters: Standard Kalman Filter, Extended Kalman Filter, Unscented Kalman Filter, and Adaptive Unscented Kalman Filter. The implementation of the lane change trajectory generator, MPC algorithm, and state estimators within MATLAB/Simulink, validated through IPG CarMaker, highlights the Adaptive Unscented Kalman Filter as the optimal choice for lane change state estimation. Its adaptive covariance adjustment sets it apart from the other filters under examination.

The security of Industrial Supply Chain (ISC) has emerged through the integration of Industrial Internet of Things (IIoT) and Blockchain (BC) technology. This new era involves effectively protecting IIoT systems from various threats and ensuring their smooth operation and resilience against potential cyber-attacks. Within the ISC ecosystem, combining machine learning (ML)-based security models for cyber-attack detection can play a crucial role in enhancing the ISC security and proactively identifying potential threats. This paper presents a BC-enabled ISC that embed ML security model integrated within a multi-layered approach. We conducted a comparative study and performance analysis of several ML classification techniques, with a focus on supervised methods to identify the lightweight model for cyber-attack detection suitable for deployment in resource-constrained IIoT environment. We investigate the performance of Gaussian Naive Bayes (NB), K-Nearest Neighbors (KNN), Random Forest (RF), Decision Tree (DT), and three ensemble techniques, namely Bagging, Stacking, and Boosting. The study employs the WUSTL-IIOT-2021 imbalanced dataset, which contains samples representing four types of attacks, including denial of service (DoS), SQL injection, reconnaissance, and backdoor. The paper addresses the imbalance in class representation by customizing the dataset for training and testing the ML models. Both Mutual Information (MI) and Extra-trees (ET) are applied as a one-stage ensemble feature selection. The performance of the ML models are investigated using classification accuracy (Acc), precision, recall, F1 score, Matthews correlation coefficient (MCC), model size (Mem), training time (TT) and prediction time (PT).

This paper introduces a novel design methodology for a dual-band branch-line coupler (DBBLC) that, for the first time, facilitates practically unlimited band ratio, enhanced flexibility in power division, and arbitrary port termination impedance concurrently. This approach ensures precise power distribution, matching, and isolation requirements by utilizing a generalized coupler core paired with an L-section impedance-matching network. This paper details an innovative and comprehensive analytical strategy for DBBLC design, which overcomes the limitations noted in prior research by deriving a generalized formula for the power division ratio (k) and simplifying the design equations to decrease complexity. This method enables the simultaneous realization of varied power division ratios, frequency ratios (r), and port impedances ( $Z_{p}$ ), thus offering remarkable design versatility. The effectiveness of this new analytical design methodology is corroborated through several design examples. Moreover, two prototype models operating at 1 GHz/2.5 GHz ( $r=2.5, k=0$ dB) and 1 GHz/2 GHz ( $r=2, k=4.77$ dB) frequencies, constructed on Rogers’ RO4003C substrate, exhibit >22 dB return loss, 25 dB isolation at all the targeted frequencies. Therefore, the development and validation of this new DBBLC structure, as demonstrated by the strong correlation between our simulated and experimental findings, not only surpasses the capabilities of existing models, but also broadens the applicability of dual-band couplers in modern wireless communication systems.

Coalition cooperative investment behavior and power allocation mechanism are key issues in the study of shared energy storage station (SESS). This paper proposes an effective alliance investment and allocation strategy to incentivize charging station operators (CSO) to invest in SESS construction. Firstly, to address the high cost problem of SESS, the paper suggests utilizing retired electric vehicle batterie (RBEV) as energy storage devices, aiming to reduce operational expenses. Additionally, a new measure called the “cooperation factor” (CF) is introduced in this paper. The CF effectively describes the complex facility scenarios and cooperation behavior among CSO and resolves the numerical unknowns related to the stability of the original coalition. Furthermore, considering the stability of CSO’s cooperation and the role of incentive investment for CSO, a two-stage investment behavior-based approach for efficient allocation of electrical energy is proposed. Finally, real data from an urban area of a city is utilized in an experiment to validate the proposed allocation strategy. The results demonstrate that the proposed strategy improves the total system return by approximately 40%, cooperative participation rate of 93.48 % and shortened payback period of 4 years. Effectively addressing the high cost problem of SESS. Moreover, compared to other allocation methods based on cooperative games, the proposed allocation method in this paper is more reasonable for promoting cooperation and highly relevant to this specific scenario.

With increasing realised traffic on transport networks, greenhouse gas emissions show a similar trend. Reducing them is a modern aspiration, creating a better place to live and moving towards sustainability. Expanding the infrastructure is often not an appropriate solution, as the system would only be fully utilised at peak times, while at less frequent times it would not even approach capacity and would require huge investment costs. An alternative to further construction work is the implementation of intelligent traffic systems, where smoother flows can achieve higher capacity by reducing the variability in the system. In a motorway environment, a common approach is Variable Speed Limit Control, where the road is divided into zones and individual speed limits are used to increase or decrease the load on the cells. This paper proposes a solution in which individual cells make decisions cooperatively, in contrast to classical state machine-based methods. Thanks to the jointly formulated goal of the agents, a predictive control method is created that leads to a reduction in emissions due to avoided shock waves and reduced waiting times. This paper presents a solution that provides a universal solution across multiple application lengths, illustrating the power of deep learning. https://github.com/istvan-knab/Variable-speed-limit-control.

To meet the needs of embedded devices for model lightweight and high-precision recognition, this paper proposes a lightweight YOLO-DLHS-P model for pig behavior recognition based on the improved YOLOv8n model. Firstly, the C2f-DRB structure is introduced at the Backbone position, and the sizeable convolutional kernel is used to extend the receptive field to enhance the spatial perception ability of the model, and to enhance the network’s ability to capture spatial information while maintaining the number of learnable parameters and computational efficiency; The LSKA attention mechanism is then introduced to be integrated into the SPPF module to construct the SPPF-LSKA structure, which significantly improves the ability of the SPPF module to aggregate features at multiple scales; Then, the downsampling at the Neck position is optimised to the HWD algorithm, which reduces the spatial resolution of the feature map while retaining more useful information and reduces the uncertainty of the information compared with the downsampling method of the baseline model; finally, the Shape-IoU is used to replace the original CIoU, which significantly improves the detection efficiency and accuracy of the model without increasing the extra computational burden. After constructing the improved YOLO-DLHS model, the improved model is then pruned using the LAMP pruning scoring algorithm to obtain a lightweight YOLO-DLHS-P model. The experimental results show that the YOLO-DLHS model improves P, mAP@0.5, and mAP@0.5-0.95 by 4.39%, 1.68%, and 3.97%, respectively, compared to the YOLOv8n model. The YOLO-DLHS-P model improves P, mAP@0.5, and mAP@0.5-0.95 by 3.37%, 1.16%, and 2.11%, and the number of parameters, computation, and model occupancy are substantially reduced by 52.49%, 54.32%, and 49.33%, respectively. Moreover, the FPS of the YOLO-DLHS-P model reaches 79 frames, which has good real-time performance for pig behavior recognition. Therefore, the improved YOLO-DLHS-P in this paper is able to reduce the demand for hardware at the time of deployment under the premise of guaranteed accuracy and provides a lightweight behavioral recognition solution for the intelligent farming of captive pigs.

This paper proposes a new space vector pulse-width modulation technique for reducing the common-mode voltage (SVPWM2) in a three-level T-type neutral-point-clamped voltage source inverter working under switch-open-circuit fault on a neutral-point-connected switch. Under this circumstance, the space vector diagram of the inverter becomes asymmetrical so that if the conventional reduced common-mode voltage space vector PWM (SVPWM1) patterns are still applied to IGBT drivers, the output waveforms are heavily deformed, thus leading to deteriorated harmonic distortion. Regarding SOC fault tolerant control, there are several traditional methods available. However, they are either costly due to additional hardware expenses or suffer from poor PMW performances using only two voltage level (2L) switching techniques. To address these issues, this paper introduces a cost-effective strategy: a reduced CMV space vector PWM strategy for fault tolerant control. This approach utilizes a hybrid two- and three-level switching principle to enhance PWM performances, resulting in reduced output harmonic distortion, lower voltage stress, and lower switching loss compared to the economical 2L PWM method. The SVPWM2 algorithm is firstly analyzed and synthesized in the $\alpha $ - $\beta $ coordinates and then implemented in MATLAB/Simulink and experimental hardware with TMS320F28377 microcontroller. Simulation and experiment waveforms are further evaluated in terms of total harmonic distortion and weighted total harmonic distortion to validate the effectiveness of the proposed algorithm.

With the popularity of electronic devices, the number of malware has increased dramatically, posing a serious threat to the digital world. Accurately identifying malware has become a research focus. However, there are many difficulties in the research, such as insufficient algorithm generalization ability, unbalanced datasets, and long processing and identification times. To address these problems, this study proposes a malware detection framework (VBDN) based on a convolutional neural network (CNN). The framework incorporates data visualization, balanced adoption, data augmentation, and convolutional neural network techniques to achieve over 90% accuracy in classifying malware on all four open-source datasets. The experimental work has two other contributions: first, it not only focuses on the overall recognition effect of the algorithm during the research process, but also on the recognition effect of each category with the help of a confusion matrix, which provides useful information for cybersecurity personnel, researchers, and others to carry out subsequent targeted research. Secondly, the balanced approach adopted in this paper has the following advantages: no need to construct a new dataset, consumes fewer hardware resources, automatically evaluates the sampling weights, etc. Additionally, to enhance the generalization ability of the algorithm and alleviate the overfitting problem, this paper employs data augmentation techniques to improve the adopted method. By comparing with several state-of-the-art algorithms, it can be observed that the VBDN framework proposed in this paper achieves the desired results in time with acceptable accuracy.

Sudoku is a popular logic-based puzzle game. Difficulty rating of a Sudoku puzzle is a difficult problem and a large number of research works have been proposed to tackle the problem. How to evaluate a difficulty rating algorithm is also a challenging problem. Different existing algorithms evaluate their effectiveness by using different datasets. Furthermore, some datasets are not based on the records played by human players. Therefore, the algorithms cannot be compared by using a standard dataset. In this paper, we construct a dataset of Sudoku puzzles with difficulty metrics experienced by human players. We first developed an Android application named Cloud Sudoku to collect the playing records. The development of Cloud Sudoku application has been briefly described in our previous work. This paper describes more details about database structures and features to collect the data. Two difficulty metrics are then proposed for each Sudoku puzzle based on the duration spent and the ratio which is successfully solved. A dataset of Sudoku puzzles with their difficulty metrics is designed and constructed. The difficulty metrics are shown to be close to the records played by human players. Therefore, the proposed dataset can be used to evaluate the difficulty rating algorithms. The dataset is now available on Github platform.

Reinforcement learning (RL) has seen an uptick in research interest in recent years, with many papers published in a plethora of different fields, topics and applications. A lot of that can be attributed to the recent advancements in machine learning (ML) and deep learning (DL) as a whole, the power of deep neural networks and the incorporation of them into reinforcement learning algorithms and techniques. This marriage of deep neural networks and traditional reinforcement learning methods has enabled reinforcement learning algorithms to achieve incredible results in use-cases that were previously deemed to be too complex. One of these highly complex scenarios is the realm of flight. In the last few years the number of papers published in this field has increased dramatically, but there is a fairly big divide between two major categories (quadcopters vs. fixed-wing aircraft). There is published work focusing on reinforcement learning using quadcopters, also known as drones, and there is published work focusing on fixed-wing aircraft, the more traditional aircraft, commonly referred to as airplanes. This paper focuses on the latter and will give an overview over the current state of reinforcement learning in the field of fixed-wing aviation. It will look into available toolkits usable for research as well as give an overview of different fields and tasks within reinforcement learning for fixed-wing aircraft control tasks. It contributes by providing detailed information about the methodologies used in each paper, while pointing out each paper’s contributions as well. A total of 48 papers on fixed-wing RL and airplane control are discussed. With the main focus lying on papers published post 2015, after the publication of Mnih et al.’s impactful deep RL paper.

With the increase in the number of Internet of Things (IoT) applications, the reliance on robust networking, reliable communications, and efficient and secure data storage is increasing day by day. Many of these applications such as cellular communications, transportation networks and Internet of Medical Things generate huge volumes of data that need to be stored and analyzed to generate insights and make decisions related to applications. Security of data storage is a paramount concern, particularly in the current cyber security threat landscape. It is important to design novel techniques and algorithms to protect the data against these upcoming attacks. This paper provides a brief overview of security challenges faced by the data storage mechanisms. In comparison to the existing review papers in this area, the work in this paper identifies three major security techniques for data storage and conducts a comprehensive review of the recent advancements related to these techniques. These techniques include Blockchain algorithms, encryption mechanisms, and deduplication algorithms for data storage. Finally, the work highlights future opportunities and challenges related to secure data storage in future networks.

In power quality applications, the long-term reliability, low maintenance cost, and low volume and weight of a STATCOM converter are of the paramount importance. The use of the series multi-level power converter topology can significantly improve both volumetric and gravimetric density as well as power conversion efficiency of the STATCOM converter. This paper presents a solution that utilizes the Novel 5 Level E Type topology for the 3-phase, 4-wire STATCOM power quality applications. A comprehensive theoretical analysis and step-by-step design process are presented and discussed in the paper. A proof of the concept prototype (12kVA $3\times 400.230$ V) has been designed, manufactured, and intensively tested and experimentally verified. The experimental results are discussed in detail and compared with state-of-the-art solutions from both academic and industrial sectors. The achieved improvements in volumetric and gravimetric density, along with ultra-high power conversion efficiency and reduced cooling effort, justifies significant benefits of the proposed topology and approach.

To handle the low frequency oscillations of the three-phase four-wire inverter grid-connected system caused by small phase-locked range, slow locking speed and large steady-state error of the traditional phase-locked loop, an improved phase-locked loop is proposed to suppress low frequency oscillation. Firstly, the paper established a sequence impedance model of three-phase four-wire inverter, and analyzed the impact of the phase-locked loop on low frequency stability in a weak power grid and the mechanism of low frequency oscillations. Secondly, based on the cross-coupling method of the Complex Filter Matrix (CFM)-Orthogonal Signal Generator (OSG), the paper combined a first-order complex filter with a positive-negative sequence voltage component relationship matrix to form an improved phase-locked loop. The improved phase-locked loop is used to reshape the positive sequence, negative sequence and zero sequence impedance of the inverter, which not only increased the impedance amplitude of the inverter system, but also realized low frequency oscillations suppression. Finally, applying the simulation and experimentation, the paper verified the validity of the established sequence impedance and the effectiveness of the proposed improved phase-locked loop to suppress low frequency oscillation in the three-phase four-wire weak power grid.

The off-line discharge between the pantograph and catenary (PAC) occurs frequently in the actual operation of an electric locomotive, which affects the quality of the received current and the service life of the pantograph. The current experiments on the off-line discharge between the PAC are mainly low-power ones and the models established are static. To study the electrical behaviors during the PAC relative motion under the actual working conditions, in this paper, the voltage and current of the contact wire (CW) during its motion are measured. And the electrical behaviors during the rise, fall and reciprocating motions of the pantograph are investigated by combining a finite element method (FEM) model based on the magnetohydrodynamics with the auto transformer (AT) traction network circuit and moving mesh method. It is found that the harmonic frequencies are mainly in the range of 0-30 MHz. The voltage distortion is more obvious during the fall of the pantograph. And the negative polarity spike pulse has a larger amplitude and occurs more frequently. The novelty of this paper lies in the experimental test with the high current and high power, as well as the systematic analysis of the electrical behaviors and multi-physical field evolution during the motion of the pantograph under the working conditions. Its contribution is to present an effective method to study the dynamic behaviors of the PAC. The research results can provide an important reference for the identification of the electrical characteristics of the PAC and the protection of the PAC equipment.

With the wide adoption of cryptocurrency, blockchain technologies have become the foundation of such digital currencies. However, this adoption has been accompanied by a surge in cryptocurrency fraud, causing significant losses to financial organizations and individuals. One way to mitigate these losses is to use Federated Learning (FL) techniques to detect fraudulent cryptocurrency transactions. This paper provides an overview of secure, privacy-preserving, and scalable Blockchain-based Federated Learning (BCFL) as a promising solution for slowing the exponential growth of cryptocurrency fraud. BCFL enables multiple entities to collaboratively train machine learning models for detecting fraudulent cryptocurrency transactions without sharing their private data, thus preserving privacy. However, Integrating differential privacy and Secure Multi-party computation (SMPC) models in BCFL presents an additional scalability challenge. This study provides an overview of BCFL, evaluating existing research on its security, privacy, and scalability challenges in detecting cryptocurrency fraud. The review explores existing research and various methodologies, highlighting advancements and challenges in creating effective, privacy-conscious fraud detection solutions for cryptocurrency transactions. We first discuss the current state of BCFL in fraud detection, along with its potential advantages and limitations, and then discuss the existing research gaps. In particular, this paper examines various BCFL frameworks, consensus algorithms, and block architectures, emphasizing their strengths and limitations in the context of cryptocurrency fraud detection to develop scalable and privacy-preserving solutions. We compare various solutions that address scalability and privacy challenges in BCFL, including adopting a geographically distributed cloud computing model that utilizes SMPC and lightweight consensus algorithms and protocols to manage computational overheads.

Recognizing anomalies is an extremely important process in data analysis, aimed at identifying patterns in data that deviate from known norms or typical standards. These anomalies are often indicative of significant, and sometimes critical, issues such as fraud, network intrusions, and system failures. Traditional anomaly detection algorithms primarily focus on the attributes of individual observations within a dataset, typically establishing a ‘normal’ profile and flagging deviations from this profile as anomalies. This paper introduces an innovative enhancement to the Isolation Forest algorithm, a renowned method for anomaly detection known for its effectiveness and efficiency, especially in large datasets. The Isolation Forest algorithm operates by randomly partitioning the data space and constructing a binary tree, where the oddity score of a data point is ascertained based on its separation from the extremity to the base of the structure, enabling the autonomous detection of outliers in a completely unsupervised manner. The methodology presented in the paper is based on repeatedly building Isolation Forest models on datasets from which individual attributes are excluded. In our research, we used the SHAP (SHapley Additive exPlanations) method which comes from game theory and is used to determine the impact of individual features on the result of the model. When training the Isolation Forest on the full dataset, the SHAP method is used to obtain the coefficients of influence of model attributes on the prediction result. Both negative and positive influences are considered significant when counting the anomaly score. On the foundations of the results from all sub-models, a weighted average is calculated, to which weights are calculated based on the SHAP model. The comparative analysis of evaluation metrics revealed a substantial enhancement attributed to the implemented methodology. The metrics used for evaluation have shown improvement in most cases from 3.5 to 6 percent point. One of the metrics have shown an improvement of 12 percent. Obtained results demonstrate that this integrated approach not only enhances the prediction accuracy of the Isolation Forest algorithm but also offers a more interpretable understanding of the data. This advancement in anomaly detection methodology promises significant implications for various fields where quick and accurate detection of outliers is paramount.

The Willow Catkin Optimization Algorithm (WCO) is a newly proposed meta-heuristic algorithm in recent years that has a simple structure and excellent optimization searching ability, but the WCO algorithm could benefit from improvements in both convergence speed and solution diversity. In this paper, the parallel technology is introduced into the WCO algorithm, and by proposing two new communication strategies, the Random Mean (RM) method and the Optimal Flight (OF) method, the goal is to utilize all solution information obtained by each subpopulation in the parallel strategy to enhance the algorithm’s performance. Additionally, the WCO algorithm has been hybridized with the Differential Evolution Algorithm (DE), and a mutation mechanism has been introduced to improve the diversity of solutions. The resulting algorithm is called the Hybrid Parallel Willow Catkin Optimization Algorithm (HPWCO). In this paper, the HPWCO algorithm is tested on the CEC2017 benchmark function set and applied to five real-world engineering optimization problems with constraints, and the experimental results were compared with three types of algorithms: the classical algorithm, the newly proposed algorithm, and the parallel algorithm. The results indicate that the HPWCO performs excellently.

As well-known, the paradigm of federated learning (FL) operates on the principle that without centralizing data into a single server, server only trains and updates global models based on the local model from multiple clients. Compared with traditional machine learning, FL enables that data availability and invisibility and preserves the data security. However, in the process of FL, recently gradient inference attack can be launched by malicious attacker to grasp gradient information, and then infer sensitive information by analyzing this gradient information. By analyzing existing schemes for combating gradient inference attacks, it can be seen that the resistance effect is not good. Designing an effective method for resisting gradient inference attacks is a challenge. In this paper, a privacy-enhanced federated learning method is proposed to efficiently defend against gradient inference attacks and improve the accuracy of the training model. In this method, the Gaussian noise has been used to enhance the training model’s ability to resist gradient inference attacks, and the Adam gradient descent method has been applied to enhance the accuracy of the training model. Finally, this paper conducted experiments on the CIFAR-10 dataset and MNIST dataset, and the experimental results showed that compared to alternatives, the proposed method has stronger resistance to gradient inference attacks and higher model accuracy.