Search

The semiconductor industry has adopted heterogeneous integration (HI), incorporating modular intellectual property (IP) blocks (chiplets) into a unified system-in-package (SiP) to overcome the slowdown in Moore’s Law and Dennard scaling and to respond to the increasing demand for advanced integrated circuits (ICs). Despite the manifold benefits of HI, such as enhanced performance, reduced area overhead, and improved yield, this transformation has also led to security vulnerabilities in the SiP supply chain and in-field operations, ranging from chiplet piracy and SiP reverse engineering (RE) to information leakage. Although conventional countermeasures provide the desired robustness for monolithic ICs, they are insufficient for addressing these challenges in the context of HI. To address these concerns, this paper presents a novel root-of-trust architecture, augmenting the process of integration using a centralized chiplet hardware security module (CHSM), aiming to provide comprehensive and robust protection throughout the SiP supply chain and in-field operations. Also, the proposed architecture equipped with the CHSM effectively addresses potential security breaches while providing robust protection against zero-day attacks through its reconfigurable capabilities. Throughout five detailed case studies, this paper performs a comprehensive security analysis to illustrate the resilience of CHSM against contemporary attack scenarios in the HI domain.

Single-channel speech enhancement is important for modern communication systems and has received a lot of attention. A convolutional neural network (CNN) successfully learns feature representations from speech spectrograms but loses spatial information due to distortion, which is important for humans to understand speech. Speech feature learning is an important ongoing research to capture higher-level representations of speech that go beyond conventional techniques. By considering the hierarchical structure and temporal relationships within speech signals, capsule networks (CapsNets) have the potential to provide more expressive and context-aware feature representations. By considering the advantages of CapNets over CNN, this study presents a model for monaural speech enhancement that keeps spatial information in a capsule and uses dynamic routing to pass it to higher layers. Dynamic routing replaces the pooling recurrent hidden states to get speech features from the outputs of the capsule. Leveraging long-term contexts provides identification of the target speaker. Therefore, a gated recurrent layer, gated recurrent unit (GRU), or long-short-term memory (LSTM), is placed above the CNN module and next to the capsule module in the architecture. This makes it viable to extract spatial features and long-term temporal dynamics. The suggested convolutional recurrent CapNet performs better compared to the models based on CNNs and recurrent neural networks. The suggested speech enhancement produces considerably better speech quality and intelligibility. With the LibriSpeech and VoiceBank+DEMAND databases, the suggested speech enhancement improves the intelligibility and quality by 18.33% and (0.94) 36.82% over the noisy mixtures.

A remote healthcare system allows patients to be monitored and diagnosed at home; however, a reliable wireless link is essential to make it possible. This study proposes a miniaturized implantable antenna resonating at 401–406 MHz, medical implant communication service band. The proposed implantable antenna has a compact 12 mm $\times$ 12 mm $\times$ 0.64 mm (92.16 mm3) footprint and is fabricated on a bio-compatible Roger R03010 material. To improve the safety, a superstrate and silicone coating around the antenna are employed to completely insulate the antenna from the surrounding tissues. The finite difference time domain and finite-element methods were utilized for numerical analysis and optimization of the proposed antenna, followed by fabrication and testing. For better validation, simulations of the proposed implantable antenna, including reflection coefficient and gain were performed within a three-layer body phantom while the measurements were conducted inside minced pork. The developed prototype of the implantable antenna exhibits 74 MHz measured bandwidth and −31.6 dBi measured gain at 402 MHz. Further, the specific absorption rate analyses were performed for safety concerns and found to be within the standard limits. Both the simulated and measured results exhibit good agreements, with the overall performance reaching the state-of-the-art.

The Internet of Things (IoT) has dramatically changed human context with the environment, ensuring productivity, comfort, and quality of life through a variety of services and applications. Nevertheless, the rapid growth of IoT devices has introduced significant security concerns like device vulnerabilities, unauthorized access, and potential data breaches.This article deals with an immediate call to empower IoT resilience against a wide range of sophisticated and prevalent cybersecurity threats. We developed two novel hybrid deep learning mechanisms, CNN-GRU (Convolutional Gated Recurrent Neural Networks) and CNN-LSTM (Convolutional Long Short-Term Memory Neural Networks), and extensively evaluated their performance on the state-of-the-art Kitsune and TON-IoT publicly available datasets. These benchmark datasets contain a variety of multivariate IoT attacks. The aim is to demonstrate the robustness of the proposed algorithms in effectively identifying telnet, password, distributed denial of service (DDoS), injection, and backdoor vulnerabilities in IoT ecosystems. We achieved approximately 99.6% accuracy in correctly distinguishing between malevolent and non-malicious activities on the Kitsune dataset. Additionally, the TON-IoT dataset demonstrated a remarkable accuracy rate of 99.00%, with minimal drops and low false alert rates. The time efficiency of both proposed algorithms renders them well-suited for deployment in IoT ecosystems. We evaluated and cross validated the proposed techniques with current benchmarks. Consequently, the proposed hybrid deep learning anomaly detection approaches not only enhance IoT security but also provide a robust control system for addressing emerging multivariate cyber threats.

Fractional order calculus (FOC) uses arbitrary order operations in differentiation and integration. Previously, the absence of methods to solve fractional differential equations limited the application of fractional order calculus. However, with the development of new methods, FOC is now applicable in various fields and has been utilized to enhance control system performance in traditional techniques. FOC is used in this study to develop a new fractional structure for the sliding mode control approach that offers improved performance and more robustness to external disturbances. The fractional order sliding mode control (FOSMC) strategy is designed using a Lyapunov-based sliding condition to ensure system stability. To enhance performance, genetic algorithms are used to adjust the fractional orders and controller parameters. The proposed fractional sliding mode control system is applied on an unmanned aerial vehicle system affected by external disturbances and then it is compared with a conventional integer order sliding mode control (IOSMC) system. Simulation results have proved the efficiency of the proposed fractional order controller where it outperforms the conventional sliding mode controller in terms of better transient dynamics and more robustness to external disturbances.

New Radio Unlicensed (NR-U) is the key representative access technology beyond 5G implementation to alleviate the spectrum crunch. NR-U shares a 5 GHz unlicensed band with WiFi, which has contention challenges for the coexisting systems due to physical and link layer protocols disparity. Being a scheduled access system, NR-U transmissions can only start at strict periodic time slots, which requires introducing a synchronization gap period in the listen-before-talk (LBT) approach. In this paper, we address these issues and analyze the impact of various gap-based NR-U approaches to the fair and efficient coexistence of the two networks. The dependency of successful spectrum access of the two systems on the gap period is also investigated. We also present a machine learning data-driven approach to unlicensed channel selection for spectrum sharing by NR-U. The results based on actual data collected from real-life WiFi deployment scenarios indicate significant improvement in coexistence performance and spectrum utilization of the unlicensed band with the proposed approach. It is shown through simulation results that the gap period before the backoff procedure provides better coexistence performance compared to the gap-based approach, where the synchronization gap is introduced after the LBT backoff. Further, the results indicate that if the gap interval exceeds a certain threshold value for each coexistence scenario, the WiFi network starts dominating the unlicensed channel, completely blocking the NR-U transmissions.

In this contribution the authors propose a hybrid Boundary Element Method – Physics Informed Neural Networks (BEM – PINN) approach, to be used for the resolution of partial differential equations arising when formulating boundary-value problems in electromagnetism. The approach retains both the advantages of integral methods (compact representation and no need to mesh large domains) and differential methods, where the term “differential” refers here to the Automatic Differentiation carried out during the training phase of the PINN. The method is easy to implement and adds an additional flexibility to purely PINN based solution methods.

Multi-agent systems (MAS) are composed of multiple autonomous agents that interact and collaborate to achieve common objectives. These systems are characterized by their decentralization, scalability, robustness, and the local perspectives of individual agents. In the realm of quadrotor applications, MAS can perform complex tasks such as formation flying, surveillance, search and rescue, and cooperative payload transport. The control of such systems often involves techniques like trajectory planning, consensus algorithms, formation control, and obstacle avoidance. This study introduces a multi-agent quadrotor system controlled using sliding mode backstepping control (SMBC), which combines the advantages of backstepping and sliding mode control to ensure stability and robustness. The novel aspect of this research is the transformation of the leader quadrotor into a fractional-order system (FOS) while maintaining the followers as integer-order systems, all controlled using the same SMBC method. MATLAB simulations validate that the fractional-order leader (FOL) achieves superior trajectory tracking compared to the integer-order leader. Furthermore, the followers tracking the fractional-order leader exhibit enhanced performance, with reduced tracking errors and quicker response times. These findings demonstrate the potential advantages of incorporating fractional-order dynamics into multi-agent control systems.

The increasing integration of cyber-physical systems (CPS) in critical infrastructures has heightened the importance of ensuring their security against various cyber threats. Game theory has emerged as a powerful analytical tool to model and analyze the strategic interactions between attackers and defenders in CPS. This survey provides a comprehensive review of the state-of-the-art research on the application of game theory in the security of CPS, with a specific focus on attack models and security analysis. We employ machine learning algorithms to classify existing research papers based on the attack target and the types of attacks they address. Our classification reveals significant trends and gaps in the current literature, offering insights into the effectiveness of different game-theoretic approaches and the prevalence of various attack models. By synthesizing the findings from over 800 research papers, we highlight the strengths and limitations of existing methodologies and propose directions for future research. This survey aims to serve as a valuable resource for researchers and practitioners seeking to enhance the security of CPS through game-theoretic frameworks and machine learning techniques.

Electroencephalogram (EEG) data annotation demands considerable expertise and is a time-intensive process. Moreover, inter-subject variability intensifies the challenge of domain shift, adversely impacting the generalization performance of deep learning models on unseen subjects. Current methods in EEG data analysis often struggle to handle the complex nature of brain activity without relying on EEG feature engineering. In this paper, we present a hybrid semi-supervised framework for seizure type classification, which relies on minimal domain knowledge provided by exploiting spectral and spatial patch-level representations of raw unlabeled EEG data, while leveraging a small amount of labeled data. Our method, IMPRESS, enhances EEG representation learning by combining multi-patch mutual information maximization with adversarial distribution alignment. We assessed the framework’s performance for cross-patient seizure classification using publicly accessible Temple University Seizure Corpus. IMPRESS surpasses the best-performing semi-supervised learning method by 1.92% and 0.72% using balanced accuracy and macro-F1 metrics, respectively, with 40 labeled samples per class. Remarkably, IMPRESS surpasses the fully-supervised method while requiring only 25 labeled samples per class. Additionally, we visualize the learned feature embeddings, highlighting the underlying dynamics across different seizure types, aiding in understanding the model’s behavior. This demonstrates the potential of leveraging multi-patch information from unlabeled data through a contrastive data-driven approach, alleviating the burden of annotating large amounts of EEG data.

The recent rise of Artificial Intelligence (AI) has already revolutionized human lives and is improving the quality of human life in many ways. The field of AI, Natural Language Processing (NLP), helps to understand, comprehend and even generate new content. NLP is used for various content analysis tasks such as sentiment analysis, fake news detection, etc. However, human personality traits detection is a new research domain. Analyzing users generated content has a significant role in identifying and understanding users’ views and behaviors. Users’ traits detection can be helpful in analysis of consumers’ personalization, finding top candidates for recruitment, career counselling, etc. In this research study, our aim is to predict personality of extroversion behaviors using machine and deep learning approaches. Extroversion means whether a person is an introvert or extrovert as this trait is relevant to certain jobs like team management, social ties etc. For empirical analysis, we investigate MBTI dataset with various feature engineering techniques including textual features like Term Frequency-Inverse Document Frequency (TF-IDF), Parts of Speech (PoS) tagging, as well as deep word embeddings ok word2vec, GloVe. The state-of-the-art shallow machine learning, ensemble modelling and deep learning models are applied. The main novelty is the exploration of latest sentence embeddings which captures semantic information of content in a better manner. Thus, the comprehensive results analysis reveals sentence embeddings as features to Bi-LSTM achieves highest accuracy of 92.52% and outperforms the existing studies in the relevant literature.

Tons of handwritten document images, either historical or continuously reproduced in everyday life, are information-rich, awaiting extensive advanced technological attention from academic and industrial communities. Document layout analysis (DLA) is a primary key that opens many doors for further sophisticated document image processing and understanding. It has been a primary interest of massive research on various domains, including document image retrieval (DIR), whereas little attention was paid to valuable semantic aspects. Recently, semantic DLA (SDLA) emerged to enable semantic information derivation and more invariant characterization. The viability of comparative-based SDLA for inferring comparative semantic characteristics has yet to be investigated. In this research, new comparative characteristics are proposed to empower more perceptive SDLA and improve retrieval capabilities. The proposed Comparative-SDLA not only utilizes the latent potency of semanticity in effectively characterizing handwritten document image layout but also exploits the power of comparability between relative semantic characteristics for further robust and enhanced DIR. A novel methodology framework is thoroughly described, implicating pairwise comparative-based automatic image annotation, document ranking by comparative characteristic, and comparative feature extraction. Several retrieval experiments on a sizable complex handwritten document dataset are conducted with extended performance evaluation, analysis, and comparison for comparative-based methods against non-comparative counterparts, highlighting promising capabilities to extend for other practical applications.

Bronchiectasis is a chronic respiratory disorder characterized by the dilation and damage of bronchial walls due to recurrent and prolonged episodes of inflammation and infection. The diagnosis of bronchiectasis requires objective measurement of abnormal bronchial dilation. It is confirmed radiologically using a chest tomography (C.T.). scan where the pathognomonic feature is increased broncho-arterial ratio (BAR) (>0.8 in children) and, often accompanied by other features such as bronchial wall thickening (W.T.). Developing image processing-based methods facilitates quicker interpretation of the scans and enable detailed evaluations according to the lobes and segments. However, challenges, such as inclined and oblique orientations of structures, and partial volume effect, can complicate accurate measurements in the upper and middle lobes using the same algorithms. Therefore, the accurate detection and measurement of airway and artery regions for BAR and wall thickness in each lobe require specialized image processing/machine learning methods and approaches. Here, we adopt a step-by-step approach and propose methods for three steps: 1. Separating right lower lobe (RLL) region from full-length C.T. scans using tracheal bifurcation (Carina) point as a central marker; 2. Updated technique to locate inner diameter of airways and outer diameter of arteries for BAR measurement; and 3. Measuring airway wall thickness (W.T.) by identifying the outer and inner diameter of airway boundaries (perimeter). Our analysis of 13 high resolution C.T. scans (HRCT) with varying thicknesses (0.67mm, 1mm, 2mm) demonstrates that the frame containing the tracheal bifurcation can be detected accurately in most cases, with a deviation of ±2 frames in some cases. Similarly, a Windows app is developed for measuring inner airway diameter, artery diameter, BAR, and wall thickness, allowing users to draw bounding around visible discrete B.A. pairs in the RLL region. Measurements of 10 B.A. pairs revealed accurate results comparable to those of a human reader, with deviations of ±0.10-0.15mm observed across all measurements. Additional studies and validation are necessary to consolidate inter- and intra-rater variability and enhance the proposed methods.

Speech enhancement (SE) aims to improve the quality and intelligibility of speech signals, particularly in the presence of noise or other distortions, to ensure reliable communication and robust speech recognition. Deep neural networks (DNNs) have shown remarkable success in SE due to their ability to learn complex patterns and representations from large amounts of data. However, they face limitations in handling long-term temporal sequences. Spiking neural networks and transformers inherently manage temporal data and capture fine-grained temporal patterns in speech signals. This paper proposes a model that integrates self-attention with spiking neural networks for speech enhancement. The proposed model employs a convolutional encoder-decoder architecture with a spiking transformer acting as a bottleneck network. The spiking self-attention mechanism in this framework represents features using spike-based queries, keys, and values. This approach enhances features by effectively capturing temporal dependencies and contextual relationships in speech signals. The spiking transformer is divided into two branches to capture comprehensive global dependencies across the temporal and spectral dimensions. The encoder-decoder incorporates a multi-scale feature extractor, which extracts features at various scales, enabling the model to build a comprehensive hierarchical representation. This representation significantly enhances the model’s ability to learn and process noisy speech, leading to excellent SE performance. Experiments are conducted using two publicly available benchmark datasets: WSJO-SI84 and VCTK+DEMAND. The proposed model demonstrated improved SE performance, showing significant progress with notable improvements of 33.69% in ESTOI, 1.05 in PESQ, and 11.36 dB in SDR over the noisy mixtures.

Dental health plays a pivotal role in determining the general health and quality of life of a person. Dental caries is a prevalent health problem in the world; therefore, prompt and effective dental caries treatment is necessary for individuals who require pain relief. Conventional methods of diagnosing dental diseases, like visual examination and radiographic testing, depend on qualified medical professionals and can be incredibly labor-intensive and imprecise in their diagnosis. To overcome these challenges, this study designed a deep learning (DL) model for the diagnosis of several dental conditions such as tooth decay, non-periodontal, and periodontal disease. A novel model named periodontal disease classification network (PDCNET) based on convolutional neural network (CNN) has been developed for the identification of periodontal disease using dental radiographs. Additionally, the proposed PDCNET model has been evaluated on two publicly available benchmark datasets of dental caries. To handle the imbalanced classes of the dental caries dataset, this study used the SMOTE TOMEK method to generate new synthetic samples for the minority categories to ensure the periodontal disease dataset is balanced. The proposed PDCNET model acquired a 99.79% AUC, 98.39% recall, 98.39% accuracy, 98.39% precision, and an F1-score of 98.31%. Furthermore, the performance of the proposed PDCNET model is contrasted using the six baseline pre-trained classifiers such as EfficientNet-B0 (M1), DenseNet-201 (M2), Vgg-16 (M3), Vgg-19 (M4), Inception-V3 (M5), and MobileNet (M6) in terms of many parameters. The levels of accuracy attained by M1, M2, M3, M4, M5, and M6 are 85.94%, 94.37%, 91.96%, 93.57%, 89.95%, and 94.77%, respectively. The findings showed that the PDCNET model has superior outcomes as compared to baseline models and provides significant assistance to the dentist in the diagnosis of dental disease.

Fuzzy logic controls (FLCs) have emerged as a promising solution for speed regulation in induction motor (IM) drives, offering adaptability to non-linearities, parameter variations, and external disturbances. However, conventional FLCs with fixed parameters and a huge number of rules can limit adaptiveness and increase system complexity, leading to deteriorated performance and high computational requirements. Moreover, reliance on costly encoders in traditional sensor-based IM drives introduces measurement errors and contributes toward the overall cost. To tackle these challenges, this paper proposes an integrated sensorless IM drive with a simplified self-tuning FLC (ST-FLC) and data-driven reinforcement learning (RL) for speed estimation. By employing a simplified 9-rule FLC instead of an intensive 49-rule counterpart and integrating a simple self-tuning mechanism based on mathematical equations, adaptiveness is maintained while computational overhead is reduced. Furthermore, the adoption of RL-based sensorless speed estimation eliminates reliance on encoder data, offering a cost-effective and computationally efficient alternative. Unlike conventional sensorless methods, the proposed sensorless-RL approach is data-driven and does not rely on motor parameters, leveraging a pre-trained policy for efficient speed estimation. Validation through simulation and experimentation on the dSPACE DS1104 platform demonstrates the efficacy of the proposed ST-FLC Sim 9-rule with sensorless RL. The method showcases accurate speed estimation, with simulation results comparable to standard 49-rule FLC and superior experimental performance. Significant computational time reduction is achieved with the proposed approach, resulting in a notable improvement in experimental performance metrics. Specifically, reductions of 50.5%, 20.4%, 15%, and 14.9% in settling time, current ripples, torque ripples, and current harmonics, respectively, underscore the practical benefits of the proposed integrated ST-FLC Sim 9-rule with sensorless-RL IM drive system.

Natural languages are written and spoken languages, and NLP (Natural Language Processing) is the ability of a computer program to recognize both written and spoken languages. Word Sense Disambiguation (WSD) is identified as a challenging area of research in Artificial Intelligence (AI), and Machine Translation (MT). WSD is the procedure for selecting the exact meaning of a word that has more than one meaning. This is an essential application for all-natural language processing applications. There are various knowledge-based, supervised, and unsupervised approaches to WSD process. The Naïve Bayes classifier as an example of approach supervised and unsupervised approaches is the most important method. In this paper, we emphasize on the use of the Naïve Bayes approach for text classification in WSD techniques. Bayes’ hypothesis is a probabilistic model and a reliable approach for text classification. Bayes’ hypothesis acknowledges that the occurrence of some other features is not dependent on the presence of a particular element in a class. This calculation can be used to solve multi-class prediction problems. This classifier performs better compared to the other methods of different approaches. This paper gives an itemized investigation of Naïve Bayes algorithms, which depicts its ideas, covering up Naïve Bayes,’ text characterization, traditional innocent Bayes,’ and machine learning. We have used the collocation method of feature extraction for the WSD of English sentences. Using this model we have disambiguate ambiguous English words by predicting part-of-speech inclusive of “noun,” “verb,” “adverb,” and “adjective.” This disambiguation module is an enhancement in machine translation. The system reported the performance measure of seventy-eight (78%) percent of the scale on F1-measure.

The rise of the Internet of Things (IoT), marked by unprecedented growth in connected devices, has created an insatiable demand for supplementary computational and communication resources. The integration of Unmanned aerial vehicles (UAVs) within IoT ecosystems presents a promising avenue to surmount these obstacles, offering enhanced network coverage, agile deployment capabilities, and efficient data gathering from geographically challenging locales. UAVs have been recognized as a compelling solution, offering extended coverage, flexibility, and reachability for IoT networks. Despite these benefits, UAV technology faces significant challenges, including limited energy resources, the necessity for adaptive responses to dynamic environments, and the imperative for autonomous operation to fulfill the evolving demands of IoT networks. In light of this, we introduce an innovative UAV-assisted data dissemination framework that aims to minimize the total energy expenditure, considering both the UAV and all spatially-distributed IoT devices. Our framework addresses three interconnected subproblems: device classification, device association, and path planning. For device classification, we employ two distinct types of deep reinforcement learning (DRL) agents—Double Deep Q-Network (DDQN) and Proximal Policy Optimization (PPO)—to classify devices into two tiers. To tackle device association, we propose an approach based on the nearest-neighbor heuristic to associate Tier 2 devices with a Tier 1 device. For path planning, we propose an approach that utilizes the Lin-Kernighan heuristic to plan the UAV’s path among the Tier 1 devices. We compare our method with three baseline approaches and demonstrate through simulation results that our approach significantly reduces energy consumption and offers a near-optimal solution in a fraction of the time required by brute force methods and ant colony heuristics. Consequently, our framework presents an efficient and practical alternative for energy-efficient data dissemination in UAV-assisted IoT networks.

Existing fully-supervised semantic segmentation methods have achieved good performance. However, they all rely on high-quality pixel-level labels. To minimize the annotation costs, weakly-supervised methods or semi-supervised methods are proposed. When such methods are applied to the infrared ship image segmentation, inaccurate object localization occurs, leading to poor segmentation results. In this paper, we propose an infrared ship segmentation (ISS) method based on weakly-supervised and semi-supervised learning, aiming to improve the performance of ISS by combining the advantages of two learning methods. It uses only image-level labels and a minimal number of pixel-level labels to segment different classes of infrared ships. Our proposed method includes three steps. First, we designed a dual-branch localization network based on ResNet50 to generate ship localization maps. Second, we trained a saliency network with minimal pixel-level labels and many localization maps to obtain ship saliency maps. Then, we optimized the saliency maps with conditional random fields and combined them with image-level labels to generate pixel-level pseudo-labels. Finally, we trained the segmentation network with these pixel-level pseudo-labels to obtain the final segmentation results. Experimental results on the infrared ship dataset collected on real sites indicate that the proposed method achieves 71.18% mean intersection over union, which is at most 56.72% and 8.75% higher than the state-of-the-art weakly-supervised and semi-supervised methods, respectively.

In power-constrained wireless networks, extending network lifespan can be achieved through effective energy harvesting techniques. Developing energy-efficient routing protocols in such networks is complex due to the unstable and unpredictable harvested energy. In this regard, reconfigurable Intelligent Surfaces (RIS), capable of manipulating the RF signals, offer enhanced energy harvesting, which in turn minimizes the nodes’ battery recharging time (BRT). Our study presents a novel RIS-assisted routing protocol for multi-hop energy harvesting-based wireless networks. The protocol assigns channels to each hop on the chosen path. It incorporates a rate calculation algorithm that takes into consideration the harvested power, BRT, and channel conditions. More specifically, it assigns a specific rate to every available channel, taking into account the energy available at each node, the estimated time required for battery recharging, and the link status between nodes. By selecting the highest rate per hop, the protocol enhances network throughput while considering energy limits and channel quality. The proposed approach is evaluated and compared with the benchmark, the Min hop scheme, using MATLAB simulations, demonstrating its superior performance.

Unmanned Aerial Vehicles (UAVs) have proliferated across diverse domains. However, optimal UAV operations necessitate precise and reliable navigation systems. UAVs predominantly rely on the Global Navigation Satellite System (GNSS), such as the Global Positioning System (GPS), for navigation. Nevertheless, GNSS signals are susceptible to blockage, reflection, and spoofing, introducing significant risks, including navigation loss and potential UAV loss. This research investigates cutting-edge navigation solutions, emphasizing deep learning-based visual localization approaches tailored for UAVs. Our focus is on scenarios characterized by GPS-denied environments where GPS signals may be absent or unreliable. We provide a comprehensive review of contemporary deep learning-based visual localization approaches and compare them to traditional aerial visual localization methods, such as template matching and feature matching. This comparison highlights both the potential benefits and challenges associated with these approaches. Furthermore, we systematically evaluate and classify recent deep learning-based methods based on main criteria, including model type/architecture, reference imagery, operational context, and resultant accuracy levels. Our findings underscore the substantial promise inherent in various approaches while also shedding light on their unique deployment challenges. Finally, we discuss potential research directions, to inspire further innovations and progress in this domain. The ultimate goal is to develop more accurate, dependable, and secure navigation solutions for UAVs.

Most companies nowadays are using digital platforms for the recruitment of new employees to make the hiring process easier. The rapid increase in the use of online platforms for job posting has resulted in fraudulent advertising. The scammers are making money through fraudulent job postings. Online recruitment fraud has emerged as an important issue in cybercrime. Therefore, it is necessary to detect fake job postings to get rid of online job scams. In recent studies, traditional machine learning and deep learning algorithms have been implemented to detect fake job postings; this research aims to use two transformer-based deep learning models, i.e., Bidirectional Encoder Representations from Transformers (BERT) and Robustly Optimized BERT-Pretraining Approach (RoBERTa) to detect fake job postings precisely. In this research, a novel dataset of fake job postings is proposed, formed by the combination of job postings from three different sources. Existing benchmark datasets are outdated and limited due to knowledge of specific job postings, which limits the existing models’ capability in detecting fraudulent jobs. Hence, we extend it with the latest job postings. Exploratory Data Analysis (EDA) highlights the class imbalance problem in detecting fake jobs, which tends the model to act aggressively toward the minority class. Responding to overcome this problem, the work at hand implements ten top-performing Synthetic Minority Oversampling Technique (SMOTE) variants. The models’ performances balanced by each SMOTE variant are analyzed and compared. All implemented approaches are performed competitively. However, BERT+SMOBD SMOTE achieved the highest balanced accuracy and recall of about 90%.

Postpartum depression (PPD) is becoming increasingly prevalent worldwide, often manifesting in new mothers due to a complex interplay of physical, behavioral, and emotional transformations post-childbirth. The primary aim of our research is to analyze the contributory factors leading to PPD, including familial, social, and other maternal health-related aspects, and to devise a predictive model that can accurately assess the risk of PPD. In this research, we analyzed a benchmark dataset of 1,503 entries gathered from a medical institution, where the data was compiled through questionnaires disseminated using a digital Google Forms platform. We deployed eleven advanced machine-learning algorithms for comparison. We proposed a novel MDKR model, a meta-learner designed to excel in predicting PPD. Questionnaire data is initially processed in the proposed MDKR through the decision tree, k-nearest classifier, and random forest models. Subsequently, the outputs from these models are fed into a meta-learner multi-layer perceptron for the final prediction. Compared to state-of-the-art studies, the proposed MDKR model surfaced as the most proficient, with an exemplary accuracy of 99% in detecting PPD. In addition, we have confirmed the performance using k-fold validation and tuning hyperparameters. In the comparative assessment of all the models concerning their ability to predict PPD risk levels, MDKR emerged as the superior model. This meta-learning model has significantly contributed to identifying pivotal factors influencing PPD, enhancing the predictive framework within maternal healthcare domains.

Otitis Media (OM), predominantly affecting children, is a significant global health issue, with an estimated 360 million pediatric cases yearly worldwide. OM causes mild and moderate conductive hearing loss which can be disabling for young children, particularly during the first three years of life when brain growth is rapid, resulting in poor speech and language development, poor communication skills, and increased vulnerability on entering school. OM therefore contributes to the global burden of all-cause hearing loss. This systematic review seeks to provide a comprehensive evaluation of pre-trained Artificial Intelligence (AI) models, including both classical Machine Learning (ML) and Deep Learning (DL), in the context of OM. This review proposes six research questions, and it summarizes the body of research across multiple domains, including the diversity and quantity of source material for training and testing models, including otoscopy images, videos, and tympanometry, and the methods used to assess quality and effectiveness in real-time settings. In addition, the review aims to provide insight into the impact and potential of AI in improving OM diagnosis and cast light on the existing challenges, such as model interpretability, limited medical expert involvement, and the need for knowledge discovery and unanswered questions, including the evolving landscape of OM diagnosis within this domain. The findings of this systematic review emphasize the importance of developing more interpretable AI models that incorporate both still images of the tympanic membrane and video recordings (with multiple frames) to maximize sensitivity and specificity of the model. In addition, collaboration with consumers and medical professionals in multiple specialties (general practice, pediatricians, audiologists and ear, nose, throat (ENT) surgeons) is needed to ensure applicability and confidence of these diagnostic digital support systems in real-world healthcare settings.

In recent years, unmanned aerial vehicles (UAVs) have grown in popularity for a variety of purposes, including parcel delivery, search operations for missing persons, and surveillance. However, autonomously navigating UAVs in dynamic environments is a challenging task due to the presence of moving objects like pedestrians. In addition, traditional deep reinforcement learning approaches suffer from slow learning rates in dynamic situations and they need substantial training data. To improve learning performance, the present study proposed an enhanced deep reinforcement learning approach that encompasses two distinct learning stages namely the reinforced and self-supervised. In the reinforced learning stage, the deep Q-learning network (DQN) has been implemented and trained guided by the loss in the bellmen equation. On the other hand, the self-supervised stage is responsible for fine-tuning the backbone layers of DQN and it was directed by the contrastive loss function. The main benefit of incorporating the self-supervised stage is to speed up the encoding of the input scene captured by the UAV camera. To further enhance the navigation performances, an obstacle detection model was embedded to reduce UAV collisions. For experimental analysis, we have utilized an outdoor UAV-simulated environment called Blocks. This environment contains stationary objects that mimic buildings, as well as moving pedestrians. The study undertaken indicates that the implementation of the self-supervised stage led to significant improvements in navigation performance. Specifically, the simulated UAV was able to navigate longer distances in the correct direction toward the goal point. Moreover, the conducted analysis shows a significant navigation performance as compared with other DQN-based approaches like double DQN and dueling DQN.

Indoor localization (IL) is a significant topic of study with several practical applications, particularly in the context of the Internet of Things (IoT) and smart cities. The area of IL has evolved greatly in recent years due to the introduction of numerous technologies such as WiFi, Bluetooth, cameras, and other sensors. Despite the growing interest in this field, there are numerous challenges and drawbacks that must be addressed to develop more accurate and sustainable systems for IL. This review study gives an in-depth look into IL, covering the most promising artificial intelligence-based and hybrid strategies that have shown excellent potential in overcoming some of the limitations of classic methods within IoT environments. In addition, the paper investigates the significance of high-quality datasets and evaluation metrics in the design and assessment of IL algorithms. Furthermore, this overview study emphasizes the crucial role that machine learning techniques, such as deep learning and transfer learning, play in the advancement of IL. A focus on the importance of IL and the various technologies, methods, and techniques that are being used to improve it. Finally, the survey highlights the need for continued research and development to create more accurate and scalable techniques that can be applied across a range of IoT-related industries, such as evacuation-egress routes, hazard-crime detection, smart occupancy-driven energy reduction and asset tracking and management.

In this article, we present a twin-tiered deep learning (DL) symbol detector architecture for a multi-user multiple-input multiple-output (MIMO) visible light communication (VLC) system, based on rate-splitting multiple access (RSMA). The DL architecture exploits long short-term memory (LSTM), by utilizing two identical LSTM-based deep neural networks (DNNs) to concurrently decode the common and private streams, within a multicarrier framework. Termed as VLC with RSMA and LSTM-based DNN (ViRSMALNet), this detector overcomes the limitations in successive interference cancellation (SIC) receivers, and decodes the RSMA messages from the combined signal in a single step. Extensive simulations based on Monte Carlo methods are executed to assess the symbol error rate (SER) performance of ViRSMALNet across diverse modulation schemes. The results establish that ViRSMALNet outperforms SIC-based least squares (LS) and minimum mean square error (MMSE) detectors, and closely attains the performance of the optimal maximum likelihood (ML) detector. Further, this study also investigates the influence of factors such as the number of light-emitting diodes per fixture, the number of photodetectors per user, and the DNNs’ hyperparameters on the SER performance. Moreover, the impact of the distance between the PDs on the system’s sum rate performance is studied.

In optical communication systems, the Q-factor is an important performance metric to evaluate the performance of an optical link. In this paper, a deep learning-based eye diagram analyzer is proposed to estimate the Q-factor. CNN architectures such as LeNet, Wide ResNet, and Inception-v4 are used for ON-Off Keying (OOK) and Pulse Amplitude Modulation (PAM) formats’ eye diagrams. The performance of these architectures is evaluated in terms of accuracy, Mean Squared Error (MSE), and error tolerance. This work shows that Wide ResNet demonstrates better performance in both OOK and PAM4 transmission schemes, achieving MSE values of 0.00188 and 0.00036, respectively. Additionally, it attains a high R-squared (R2) value of 0.9998. This deep learning-based eye diagram analyzer may be a promising approach for analyzing and optimizing optical communication systems without extensive human intervention.

A modern electric power system integrated with advanced technologies such as sensors and smart meters is referred to as a “smart grids”, aimed at enhancing electrical power delivery efficiency and reliability. However, fault location and prediction can become challenging when dynamic fault currents from renewable energy sources are present. To address these challenges, three unique deep learning models that make use of Deep Neural Networks (DNN) have been proposed. CNN, LSTM, and Hybrid CNN-LSTM are deep learning models. Line faulty identification (LF), fault classification (FC), and fault location estimate (FL) are the subjects on which they concentrate. These models analyze data gathered both pre and post faults occur in order to enhance decision making. Signals including the voltage and current were fed into these models from many different locations across the test networks. Once the 1D CNN has extracted characteristics from the gathered signals, LSTM uses these features to make accurate estimations and identify faults. Complex data are compatible with this method in terms of optimal outcomes. Using training and testing data from transmission line failure simulations, the proposed approaches were evaluated on the IEEE 6-bus and IEEE 9-bus systems. The tests encompassed a range of fault classes, locations, and ground fault resistances at various locations. Distributed Generator (DG) resources were additionally included in the system architecture and changes in the topology of the networks were considered in terms of location and number of DG resources. The results demonstrated that the proposed algorithms outperformed contemporary technologies in terms of detection, classification, and location accuracy. They demonstrated high accuracy and robustness in their performance.

Gas-insulated systems are widely utilized in the electric power sector to transmit and distribute electrical energy. Sulphur-hexafluoride (SF6) has dominated gas insulation in high-voltage insulation technology since the early 60s. It is a greenhouse gas with a protracted lifespan in the atmosphere. This paper proposes an economical and comparatively more environmentally friendly R507 gas alternative to SF6 for medium-voltage applications. R507 has been analyzed experimentally through power frequency breakdown and lightning impulse testing to validate the performance and theoretical concepts. R507 has a very low liquefication temperature of $-46.7^{\circ }\text{C}$ , but it must still be mixed with buffer gases such as CO2, N2, or dry air to meet the diverse liquefaction temperature applications. Various field utilization factors under AC and lightning impulse voltages are used in the experiments, along with different electrode geometries, including sphere-to-plane and rod-to-plane (i.e., quasi-homogeneous and inhomogeneous electric field distribution). For comparison, identical experiments are conducted with pure SF6. R507 gas was found to be a promising substitute for SF6 gas, with its dielectric strength being approximately 0.95 times that of SF6 gas. A positive synergistic effect is present between R507 and CO2, along with the good self-recoverability property of the gas mixture. The current research study serves as a fundamental resource for characterizing the R507/CO2 gas mixture insulation properties to be utilized in practical applications.

This study introduces a novel approach to enhance state and unknown input estimation for the synchronous reluctance motor, utilizing the Takagi-Sugeno fuzzy representation. A proportional multi-integral observer structure is introduced, capable of capturing a wider range of unknown input dynamics compared to the previous proportional integral observer utilized in prior research. Stability conditions, obtained using the quadratic Lyapunov function, are formulated as linear matrix inequalities to guarantee the asymptotic convergence of the estimation error. To evaluate the effectiveness of the proposed observer, different cases were considered, including scenarios with slow and fast variations of the unknown input. Through hardware-in-the-loop validation, the results clearly demonstrate the superiority of the proposed method over the earlier approach. This advancement signifies a significant improvement in the estimation accuracy and reliability of the Synchronous Reluctance Motor.

One of the challenges in image processing involves single image deraining (SID). Existing methods do not exploit images at multi-scales and often overlook spatial and channel information. These methods fail to account for different rain conditions, leading to challenges in effectively removing a wide range of rain streak patterns, such as diverse directional or dense streaks. Moreover, they often result in the loss of fine texture details or a blurred background in the process of eliminating rain streaks from images captured in heavy rain. Mostly state-of-the-art (SOTA) deraining models achieve higher performance in removing rain from rainy images but at the expense of a high number of parameters, which results in computational complexity and memory requirements. Furthermore, they also do not consider high-level visioned evaluation metrics to perform deep evaluations of the proposed models. In this paper, a simple lightweight network with few parameters and relatively shallow depth is proposed by fusing the traditional Gaussian-Laplacian pyramid technique with the attention module. We propose a novel attention-based lightweight pyramid network for image deraining and detection (LPN-IDD) to achieve better deraining performance. The proposed model includes a dual attention module integrated with the Gaussian-Laplacian pyramids network. In LPN-IDD, different levels of Laplacian pyramids can extract multi-scale features to adapt to different shapes and types of rain streaks. Residual and recursive blocks are used in each subnetwork along with dual attention blocks to resist the occlusion or texture feature while suppressing unnecessary features. Extensive experimentation performed on the SID synthetic and real-world datasets demonstrate the effectiveness of the proposed model in image deraining tasks. For deep evaluation of the proposed methodology, object detection models including faster-RCNN, YOLO-V3, YOLO-V7 are used along with full reference evaluation metrics.

Accents, or changes in how different people speak the same word/sentence in the same language, pose substantial communication issues in most spoken languages. This is a well-known fact, but how does the accent of one language affect learning/speaking another? In this paper, we look at how Arab accents influence the English language. To that end, we built a deep machine-learning system for Arabic accent recognition that was learned from an in-house English speech database of four Arabic accents collected from Jordan, Iraq, Saudi Arabia, and Tunisia. The proposed system employs Mel spectrograms of an English-spoken paragraph to train an LSTM neural network to recognize the accent in each sound signal. Although the collected data was extremely difficult to learn due to the presence of both males and females and fluent speakers in each class, the proposed system could recognize speakers with various accents by up to 79%. This answers the study’s main question, demonstrating that speakers with an Arabic accent have their way of speaking English, which varies by country. As a result, if trained on appropriate and adequate data, the proposed system can also be used to recognize accents in any language.

A novel automated multi-classification approach is proposed for the anticipation of lung abnormalities using chest X-ray and CT images. The study leverages a publicly accessible dataset with an insufficient and unbalanced number of images, addressing this issue by employing the data augmentation approach DCGAN to balance the dataset. Various preprocessing procedures are applied to improve features and reduce noise in lung pictures. As the base for the model, the vision trans-former and convolution-based compact convolutional transformer (CCT) model is utilized. To determine the best model configuration, an ablation study is performed on the original CCT model using a CT scan dataset with image dimensions of $32\times32$ . Following that, this model is trained on the X-ray dataset to evaluate performance on an entirely other modality. The performances are compared to six pre-trained models with $32\times 32$ images. While traditional models achieved modest performance, with test accuracies ranging from 43% to 77% and 49% to 73% requiring lengthy training times, the suggested model performed exceptionally well, obtaining test accuracies of 99.77% and 95.37% for CT and X-ray, respectively with a short training duration of 10–12 and 40–42 seconds/epoch. Robustness is demonstrated through the progressive reduction of the number of training images, with findings indicating that the model maintains good performance even on a reduced dataset. An explainable AI technique Grad-CAM is used to explain the model’s judgment. Grad-CAM-based color visualization is shown to explain model assessments and help health specialists make quick, confident decisions. This study used image preprocessing and deep learning techniques to detect lung anomalies, and it addressed the challenges of training time and computational complexity.

Large Language Models (LLMs) recently demonstrated extraordinary capability in various natural language processing (NLP) tasks including language translation, text generation, question answering, etc. Moreover, LLMs are new and essential part of computerized language processing, having the ability to understand complex verbal patterns and generate coherent and appropriate replies in a given context. Though this success of LLMs has prompted a substantial increase in research contributions, rapid growth has made it difficult to understand the overall impact of these improvements. Since a plethora of research on LLMs have been appeared within a short time, it is quite impossible to track all of these and get an overview of the current state of research in this area. Consequently, the research community would benefit from a short but thorough review of the recent changes in this area. This article thoroughly overviews LLMs, including their history, architectures, transformers, resources, training methods, applications, impacts, challenges, etc. This paper begins by discussing the fundamental concepts of LLMs with its traditional pipeline of the LLMs training phase. Then the paper provides an overview of the existing works, the history of LLMs, their evolution over time, the architecture of transformers in LLMs, the different resources of LLMs, and the different training methods that have been used to train them. The paper also demonstrates the datasets utilized in the studies. After that, the paper discusses the wide range of applications of LLMs, including biomedical and healthcare, education, social, business, and agriculture. The study also illustrates how LLMs create an impact on society and shape the future of AI and how they can be used to solve real-world problems. Finally, the paper also explores open issues and challenges to deploy LLMs in real-world scenario. Our review paper aims to help practitioners, researchers, and experts thoroughly understand the evolution of LLMs, pre-trained architectures, applications, challenges, and future goals.

In this paper, a leader-follower-based cooperative and formation flight control is designed for a group of quadrotors using new forms of fractional sliding mode control (FSMC) strategy accompanied by the potential field algorithm. The FSMC strategy is employed as an inner controller to stabilize each individual UAV quadrotor while the potential field function is used as a formation algorithm (output controller) in order to keep a desired shape formation during the navigation of the fleet. Firstly, the stability of a single quadrotor subjected to external disturbances is addressed by designing an FSMC in addition to a classical PID controller. The FSMC is responsible for stabilizing the altitude and attitude of the vehicle while the responsibility of the PID is to control the two-dimensional (x-y) position of the UAV. Lyapunov-based sliding condition is used in the design of the fractional control strategy to guarantee system stability. The controller’s parameters are tuned using the genetic algorithm (GA) to obtain better performance and more robustness against external disturbances. The potential field algorithm is used to generate the paths that are required for the fleet of vehicles to navigate while keeping a prescribed formation shape. This is done by using the attractive potential field to attract the followers toward the leader and the repulsive potential field to repulse every two neighboring followers in order to keep a required distance between them. Simulation results have proved the efficiency of the proposed control system for both the single and multi-agent cases.

Image enhancement and reconstruction is an important field of research in digital image analysis. To increase the quality of low-contrast images, a variety of image-enhancing technologies are available. Among these image enhancement techniques, singular value decomposition and discrete wavelet transform are the popular approaches for enhancing low-quality images. In this study, we have developed a modified singular value decomposition approach to enhance low-contrast and low-resolution close-circuit television images. Low-quality and low-resolution images with singular values near zero are used as input in the proposed approach. On the selected data, a threshold value was determined in the singular values of the diagonal matrix. Finally, the proposed modified singular value decomposition technique was applied to enhance the value of images with low quality and resolution. The datasets include 90 facial images of different individuals with low quality and resolution from a university database. The performance of the proposed approach is assessed using image entropy, peak signal to noise ratio, mean square error, contrast measurement, time computation, structure similarity index measurement and image enhancement factor. During the experiments performed on the obtained dataset, it was found that the proposed modified singular value decomposition approach outperformed the existing singular value decomposition, discrete wavelet transform-singular value decomposition and stationary wavelet transform methods.

Biometric recognition using retinal and iris images is a refined and extremely safe technology in the biometrics field. The retina and iris are exclusive and constant functional features of the human eye that can be employed for individual identification. Retinal and iris detection systems are highly well-known for their high exactness and safety. The individuality and constancy of these features make them challenging in order to spoof or replicate. Biometric detection utilizing retinal and iris images improved by deep learning (DL) models has accompanied a novel period of highly exact and effective identity verification. DL models namely convolutional neural networks (CNN) and recurrent neural networks (RNN) employed for feature removal as well as matching, permitting complex and individual patterns of retina and iris to be taken with notable precision. This technique provides enlarged security and reliability. It is precious in numerous applications like access control, border control, healthcare, and mobile device verification while addressing the challenges of flexible lighting situations and accommodating users with eye conditions. This article presents an effective secure biometric retinal iris identification using an orca predator’s algorithm with deep learning (SBRIC-OPADL) technique. The main aim of the SBRIC-OPADL technique is to accomplish biometric security using retinal iris images. Primarily, the SBRIC-OPADL technique exploits the Wiener filtering (WF) approach for the removal of noise that exists in the input iris images. Besides, the SBRIC-OPADL technique exploits the EfficientNet model for the extraction of feature vectors. Moreover, the hyperparameter tuning process of the EfficientNet model takes place using OPA. Furthermore, the biometric identification process can be performed by the use of a convolutional autoencoder (CAE). To validate the enhanced biometric detection results of the SBRIC-OPADL technique is tested using the biometric iris dataset. The extensive results highlighted that the SBRIC-OPADL technique reaches better performance over other models.

The snow ablation optimization (SAO) is a new metaheuristic motivated by the melting and sublimation properties of snow. In this work, the economic load dispatch (ELD) problem, one of the key components of a power system, is solved using the SAO. There is one kind of ELD, that is focused on minimizing fuel usage costs. Assessing the reliability of the SAO, its performance is compared against some techniques. For the same case study, these techniques include the grey wolf optimization (GWO), the tunicate swarm algorithm (TSA), the monarch butterfly optimization (MBO), and the rime-ice algorithm (RIME). There are six cases used in this work: the first two cases are 6 generators at two loads 700 MW and 1000 MW for the ELD problem. The second two cases are 10 generators at two loads 1000 MW and 2000 MW for the ELD problem. The third two cases are 20 generators at two loads 2000 MW and 3000 MW for the ELD problem. The methods were assessed across 30 different runs using metrics for the maximum, mean, minimum objective function, and standard deviation. The primary component of ELD issues is the power mismatch element. This factor’s optimal value must approach zero. The optimal power mismatch values of 3.336E-13 and 1.57E-10 are obtained using the SAO method for six generator units at demand loads of 1000 MW and 700 MW, respectively. The optimal power mismatch values of 6.83E-6 and 1.65E-7 are obtained by the SAO method for ten generator units at demand loads of 1000 and 2000 MW, respectively. The optimal power mismatch values of 1.82E-4 and 7.91E-5 are obtained using the SAO method for 20 generator units at demand loads of 2000 and 3000 MW, respectively. The results produced for the six ELD case studies show that the SAO surpasses all competing algorithms, proving its superiority.

In various image-processing applications, the fractional-order functions outperform their equivalent integer-order functions. In this study, we proposed a novel technique to encrypt color images with fractional-order chaotic systems, where the fractional-order simple Loranz is integrated with the FQM. Our novel part is hybridizing two maps (Lorenz & Fibonacci). The new algorithm benefits from the strengths of the fractional-order simple Lorenz chaotic system and the FQM to get a good performance technique. We confirm that our suggested encryption algorithms are effective and robust against attacks. Our suggested technique comprises numerous sequences: The primary color channels of the input image were separated into R-, G, and B-channel. The processes of confusion and diffusion are independent for each channel. The simple Loranz generates random integers using fractional ordering to enable pixel placements. We utilized the Fibonacci Q-matrix to dilute every $2^{\ast} 2$ block comprising the permitted image.

Cooperative communications is a core research area in wireless vehicular networks (WVNs), thanks to its capability to provide a certain degree of fading mitigation and to improve spectral efficiency. In a cooperative scenario, the intercept probability of the system can be reduced by optimizing the relay selection scheme in order to select the optimal relay from a set of available relays for data transmission. However, due to the mobility of WVNs, the best relay is often selected in practice based on outdated channel state information (CSI), which in turn affects the overall system performance. Therefore, there is a need for a robust relay selection scheme (RSS) that guarantees a satisfactory overall achievable performance in the presence of an outdated CSI. Motivated by this and considering the advantageous features of autoregressive moving average (ARMA), the proposed contribution models a cooperative vehicular communication scenario with relay selection as a Markov decision process (MDP) and proposes two deep Q-networks (DQNs), namely DQN-RSS and DQN-RSS-ARMA. In the proposed framework, two deep reinforcement learning (RL)-based RSS are trained based on the intercept probability, aiming to select the optimal vehicular relay from a set of multiple relays. We then compare the proposed RSS with the conventional methods and evaluate the performance of the network from the security point of view. Simulation results show that DQN-RSS and DQN-RSS-ARMA perform better than conventional approaches, as they reduce intercept probability by approximately 15% and 30%, respectively, compared to the standard ARMA approach.

Bronchiectasis, one of the most neglected chronic lung conditions, has a high individual disease burden and economic cost and causes poor quality of life in children/adolescents and adults. Advances in image quality and a dramatic reduction in acquisition times, multiple high-resolution chest tomography (HRCT) acquisitions, and reconstructions of the lung have resulted in accurate categorization and determination of the extent of lung parenchyma and airway abnormality. For bronchiectasis, the diagnosis is confirmed using the key feature of abnormally increased broncho-arterial (B.A.) ratio (BAR), with or without other abnormalities, e.g., bronchial wall thickening, lack of bronchial tapering, and mucus plugging. Most of these features require shape analysis of the airway and artery regions to perform various assessments that can have inter-rater variability and are time-consuming. This challenge is amplified in pediatric patients due to age-related anatomical variations. The anatomical differences and variations in airway structures between Infants, Early Childhood, and middle Childhood can impact how the images can be processed and analyzed. To address this, we proposed two novel image-processing methods to detect and measure the B.A. pairs. The first method uses an optimized connected component labelling (CCL) algorithm to construct bounding boxes around the objects (airway, artery) and extract the regions of interest (ROIs) for potential B.A. pairs. The second method allows us to calculate 4 or 6 diameters for each object in the ROIs and use their mean value as the final diameter, demonstrating agreement with manual readings. Evaluating against a diverse set of HRCT scans from various categories validates the significance and practical utility of our proposed methods in detecting and measuring the disjointed B.A. pairs to assess increased BAR.

Industry is increasingly adopting high-speed electric machines due to their high-power density and low cost. However, increasing the speed leads to many challenges that are worth to study since they have relevant impact on the design of the machine. One of these challenges is the losses in stator winding assembly. At low frequency, losses in windings are usually assumed equal to DC losses. However, at high frequency some phenomena become noticeable and worth evaluating since they lead to noticeable losses. AC losses in windings are due to skin effect, proximity effect, and circulating currents. The main models of AC losses in windings in the literature are presented for both circular and rectangular wires. These models are compared using both interior and surface-mounted permanent magnet synchronous machines. Results show that for proximity effect, analytical model is convenient when considering armature reaction only, however, when considering on load case, hybrid model with simple FEA is suitable thanks to its interesting trade-off between accuracy and computation time, which makes it suitable for large-scale optimization. For circulating currents effect, finite element model is unavoidable. Moreover, two important properties concerning circulating currents are stated as well as their mathematical proofs.

Cover crops are intermediary crops planted in between two main cash crops. They play a role in limiting nitrate leaching into groundwater. Currently, there is no database pertaining to cover crops despite their importance. The development period of cover crops is characterized by a dense cloud cover in Europe, which obstructs land surface monitoring using optical remote sensing. This study proposes a cover crops mapping method based on synthetic aperture radar remote sensing data from the Sentinel-1 (S1) constellation, which is unaffected by weather conditions. Our method is based on the dynamics of the S1 backscattering coefficient at the plot level. Using a decision tree, we mapped cover crops. In the decision tree algorithm, filters were added to eliminate other crops that temporally intersect with the cover crop, namely wheat and rapeseed. The proposed decision tree proved effective in detecting existing cover crop plots, as shown by the classification Recall values ranging between 83.5% and 95.0% and the high precision values ranging between 81.5% and 89.2%. Comparison with the Random Forest classifier showed that our proposed method yielded better and more consistent results. The main limitations in the classification approach were weak cover crops and residual vegetation. The results show that the developed approach, based on the S1 time series, is capable of remotely monitoring cover crops, giving managers and decision makers the ability to follow farmers’ work and ascertain if they are applying the recommended agricultural practices that promote sustainable land use and limit the contamination of groundwater.

The presence of various maize plant leaf diseases has significantly decreased both the quality and quantity of crop production. In order to take the appropriate steps to prevent the occurrence of plant leaf diseases, it is essential to track and recognize such infections during the planting period. However, correct recognition of numerous maize diseases is difficult to achieve because the currently employed automated solutions are operationally complicated or only effective on samples with plain backgrounds. While real-world scenarios are suffering from huge sample distortions like the effect of noise, clutter in the background, and blurring of the leaf regions that increase the complexity of the recognition procedure. To alleviate the above-listed problems, a deep learning (DL) approach called the MaizeNet is proposed for the correct localization and classification of various maize crop leaf disorders. We have presented an improved Faster-RCNN approach that utilizes the ResNet-50 model with spatial-channel attention as its base network for the computation of deep keypoints which are then localized and categorized into various classes. The proposed work is tested on a standard database named Corn Disease and Severity that contains samples from three different classes of maize plant diseases which are captured under diverse conditions such as complex background, brightness, and color and size variations. The MaizeNet model attains an average accuracy score of 97.89% along with mAP value of 0.94, which is showing the effectiveness of our approach for locating and classifying the numerous types of maize leaf infections.

Ensuring reliable and easily accessible charging infrastructure becomes crucial as more people adopt electric vehicles. This study introduces a recommendation system designed to assist electric vehicle users in finding convenient charging stations, enhancing the charging experience, and reducing range anxiety. The system employs advanced data analysis techniques to offer personalized suggestions based on users’ preferences. Real-time data on factors like charging station availability, individual preferences, and past usage patterns are collected and processed using a restricted Boltzmann machine-learning algorithm. The waterwheel plant algorithm, known for its effectiveness in solving complex optimization problems, is utilized to optimize the parameters of the restricted Boltzmann machine. The recommendation system considers various user preferences, including charging speed, cost, network compatibility, amenities, and proximity to the user’s current location. The system aims to minimize user frustration, improve charging performance, and enhance customer satisfaction by addressing these aspects. Results indicate the system’s efficiency in suggesting convenient charging locations. The study explores the statistical significance of the optimized waterwheel plant algorithm and restricted Boltzmann machine model through Wilcoxon rank-sum and Analysis of Variance tests.

The potholes in the road cause substantial monetary and physical damage to the ongoing traffic. Also, it requires extensive maintenance, especially in areas where the temperature may go down below freezing point. One of the major causes of potholes is the rain or running water that is accumulated in the cracks and later on due to low temperature, changes into ice. Then, ice forms a larger volume for the same amount of water after expanding and causes cracks to expand and at some point become a pothole. They are also the cause of traffic congestion. Therefore, pothole repair needs urgent attention and is one of the routine road maintenance tasks. Kansas City is one of the cities with established social data networks for residents to request road services to mitigate the problem. Although the policymakers have not ignored the issue and rudimentary patching policies are in place, unfortunately, that does not provide efficient road maintenance routes. This paper proposes a Hierarchical Clustering Algorithm for Road Service Routing Enhancement (HiCARE). HiCARE is a practical framework, that makes use of open data in order to optimize the route for maintenance. The optimization considers the time and date of reported potholes, locations, traffic situations, weather conditions, type of patch or other repairs needed, and crew availability, to mention some. This research has characterized spatiotemporal pothole density by analyzing the past 16 years of pothole data from the Open Data KC 311 in Kansas City. HiCARE enhances the NP-hard Traveling Salesperson Problem (TSP) by classifying potholes into layers of clusters. HiCARE finds a cluster’s shortest possible pothole route by identifying each cluster group’s entrance and exit pothole points. Moreover, it modifies the final routes to skip any local minima. The empirical research and analysis indicate that HiCARE significantly reduces the traversing distance and is faster in computation time when compared to typical TSP heuristic algorithms for daily resolution scheduling.

The electric power system is transforming from a strict hierarchy into a distributed scheme that consists of prosumers and other service providers. Some services require interactions more complex than request-response. To reach functional interoperability, such services necessitate an agreement about what actions happen, in which order, and in which conditions. Such agreements can be modeled as service choreographies, a concept familiar from microservice architectures (MSA). However, the management of choreographies is demanding, especially in inter-organizational (or actor-to-actor) schemes, if changes occur frequently or the number of participants and choreography instances is high. This article proposes Discografia, a framework to support the distributed execution of service choreographies modeled with the standard Business Process Model and Notation 2.0 (BPMN2). Beyond distributed execution, certain operations must be centralized, specifically for security as well as lifecycle and execution support. The framework is proven with a software prototype to demonstrate applicability. The prototype coordinates the charging of electric vehicles (EV) among competing aggregators operating in urban charging hubs (UCH). Additionally, the concept is evaluated with theoretical calculations that suggest reduced manual effort in communicating choreography specifications. Thus, this paper argues that Discografia concept for distributed service choreographing facilitates inter-organizational interoperability, an essential factor in future power system.

The measurement and assessment of academic performance is now a fact of scientific life. This assessment guides the scientific community in making significant judgments such as selecting appropriate candidates for various positions, nominating individuals for scientific awards, and awarding scholarships or grants. Several research assessment parameters have been proposed by researchers to identify the most influential scholars. In the literature, researchers have employed a combination of hypothetical and fictional scenarios, as well as manual approaches, to identify the best assessment parameters. Moreover, there is no established benchmark available for assessing these parameters. The current study employs an innovative machine learning approach, the Dynamic Random Forest with Brouta Optimizer called “BorutaRanked Forest”, to prioritize the assessment metrics for researchers by calculating the importance score for each metric. Thirty different assessment metrics have been evaluated on a comprehensive dataset of researchers that contains awardees researchers and non-awardees researchers of three decades from (1990 to 2023). The main purpose of this evaluation is to determine the potential value and significance of each parameter relative to others. In addition, the position of awardees researchers is examined at different percentile ranges form Top 10% to Top 100% in the ranked lists of each parameter. During the individual evaluation of each parameter, we uncovered several intriguing patterns in the data. Our findings indicate that the normalized h-index is a particularly effective assessment parameter for the impact evaluation of researchers in the domain of mathematics. An analysis has been conducted to explore the correlation between parameters and awarding societies, examining the associations between different metrics and specific awarding societies.

With the advent of 5G communication networks, many novel areas of research have emerged and the spectrum of communicating objects has been diversified. Network Function Virtualization (NFV), and Software Defined Networking (SDN), are the two broader areas that are tremendously being explored to optimize the network performance parameters. Cellular Vehicle-to-Everything (C-V2X) is one such example of where end-to-end communication is developed with the aid of intervening network slices. Adoption of these technologies enables a shift towards Ultra-Reliable Low-Latency Communication (URLLC) across various domains including autonomous vehicles that demand a hundred percent Quality of Service (QoS) and extremely low latency rates. Due to the limitation of resources to ensure such communication requirements, telecom operators are profoundly researching software solutions for network resource allocation optimally. The concept of Network Slicing (NS) emerged from such end-to-end network resource allocation where connecting devices are routed toward the suitable resources to meet their requirements. Nevertheless, the bias, in terms of finding the best slice, observed in the network slices renders a non-optimal distribution of resources. To cater to such issues, a Deep Learning approach has been developed in this paper. The incoming traffic has been allocated network slices based on data-driven decisions as well as predictive network analysis for the future. A Long Short Term Memory (LSTM) time series prediction approach has been adopted that renders optimal resource utilization, lower latency rates, and high reliability across the network. The model will further ensure packet prioritization and will retain resource margin for crucial ones.